Real Numbers

Real Numbers?
By Ng Tze Beng
Department of Mathematics National University of Singapore 2 Science Drive 2 Singapore 117543
Preface
When we study calculus we often take the real numbers for granted. It is perhaps difficult for the novice to trace the development of the real numbers through the more mathematically natural but conceptually difficult path. The properties of the real numbers are well known but often assumed as a matter of fact. We shall recreate the real numbers through Dedekind's cuts, starting from the rational numbers and then prove all the properties of the real numbers that we often assumed. One last step in this approach is to show that there is only one real number system, a complete totally ordered field up to isomorphism. By this isomorphism, we understand and translate all the properties that the real numbers possess as Dedekind's cuts, to our usual concept of the real numbers as 'extension' of the rational numbers. There are other models for the real numbers, but the Dedekind's cuts offer the clearest but by no means easy, logical development. This book aims to provide the novice first year undergraduates or a beginner thirsting for the meaning of the real number system, a step by step approach to finding out about the real numbers. The need for calculus students, who we can say study properties of function from the real numbers to the real numbers or from Cartesian products of the real numbers to Cartesian products of the real numbers, to know about the object they are studying is apparent. More precisely, if we are studying the properties of the functions from the real numbers to the real numbers we ought to know or at least be familiar with the real numbers themselves. The creativity evident in the construction of the Dedekind's cuts and the deep result that there is one and only one complete totally ordered field up to isomorphism can only be appreciated with some effort and concentration. The book is written to allow for appreciation in steps rather than all at one go. If one wants to know just the real numbers as cuts, start with chapter two. Then if one wants to know about addition on the real numbers, one can proceed next to chapter three. If one wants to know about the overall picture, then start with chapter one, the least technical of all the chapters. This book has come to be written for the following reason. The author often has to clarify the fine points and the understanding of the real numbers, in particular, their properties, with students, who often regard the real numbers as little more than abstract algebra with inequalities. It is the hope that this book will make some difference towards an appreciation of the definitions of limits, continuity and differentiability and theorems such as the intermediate value theorem, extreme value theorem and the mean value theorem. In writing this book, the author has considered among other things, how anyone would begin this journey to know the real numbers better, the possible pitfalls, the explanation and the technical details from the point of view of someone doing it for the first time. This book that has gone through many revisions is also a labour of love. Many people have contributed to it in one way or another. They include the students, for whom it is mainly intended, and my family without whose support and love I will not be able to find the reason for writing it. Last but not least, thanks are due to Ian Shelly who has prompted me to write and my spirit for wanting to give. Ng Tze Beng 47 Hindhede Walk #02-01 Singapore 587977
ii Ng Tze Beng
Contents
PREFACE ii 1
Chapter 1 Real Numbers ! What are they? Chapter 2 A Cut, An Instan t 12 Chapter 3 Relearning Addition Chapter 4 Rethinking Multiplication 19 25
Chapter 5 Order! Order! There is only one order. Chapter 6 Genealogy EPILOGUE REFERENCES 57 59 51
41
CHAPTER ONE Real Numbers! What are they?
Numbers have different origins, but mostly in the abstract. There are many ways of thinking and representing integers. We speak of 'real' numbers. What do we mean by a number that is real? A question, hardly understood and least to answer. Let us start by counting. ONE, TWO, THREE, FOUR, FIVE, SIX, ..... . The counting numbers as the name suggests has to do with counting. Starting with one we get its successor, two, followed by its successor, three and so on. If we were to ask "What this number (which is represented by a symbol) is?", there is a number of answers we can give. We can start by recalling the sequence of successor followed by successor, starting from one, until we reach our desired number. We are accustomed to see numbers abbreviated by universally understood symbols. We would subsequently simplify or modify this to a system that takes into account certain specified numbers together with the operation of addition (and sometimes multiplication) that gives a representation of a number by a shorthand that will let the beholder know and reconstruct the number by addition of certain basic chosen numbers. For example, 1024 would mean one THOUSAND, zero HUNDRED, two TEN and four ONE. How useful or effective is this representation? It would depend on the chosen specified blocks of numbers and how you would represent these blocks. A symbol for each block would be cumbersome. For one we would run out of symbols. For second these symbols may not be universally understood. Nevertheless, these symbols are unavoidable if we want to make ourselves understood. For instance, 1024 could equally mean one 63 , zero 62 , two 6 and four 1. There would also be the need for special symbols to represent certain numbers. These would be the first few numbers needed to exchange certain number of blocks with another block. This is useful and desirable for commerce and trade. For our standard base TEN representation, we would need the symbols for (zero), one, two, ..., nine. A symbol for zero is not essential for this purpose. Ten ONE is TEN or one block of TEN, ten blocks of TEN is HUNDRED and so on. For the base 2 representation we would need only zero and one. So two ONE is 2 , two blocks of 2 is 22 and so on. The representation ought to be unique otherwise we would be open to dispute. Because a symbol for each block is cumbersome, although there is a place for this, place value becomes a device to circumvent the need for these symbols or to reduce the number of symbols required. But then there is still the problem of which way do we assume the place value to begin, from the right to the left or from the left to the right? A symbol for zero here would be essential because we cannot disregard the place value even though there is zero block there. 1024 and 4201 could mean the same thing unless we all agree on which way the place value is written down. We shall adopt the base 10 representation from right to left starting from the smallest to the largest blocks.
Chapter 1 Real Numbers! What are they?
The Integers
To allow for the use of arithmetic, we would then introduce the notion of negative numbers. This is of course desirable for trading purposes. Mathematicians would view negative numbers as the solution to the linear equation, x+n=0 for each counting number n. For this approach, of course the notion of zero would have to be introduced. There is not a satisfactory way of defining or saying what zero means at least mathematically or what the equation means, unless we tug at the very nature of the counting numbers, the meaning of addition, multiplication and the ordering of the counting numbers and its construction together with the meaning of zero as satisfying the following: m + 0 = m for all counting numbers m and 0 + 0 = 0. So we must then assume the existence of such a number zero and rightly this is a very important number and we should indeed construct the counting numbers together with zero. We shall not venture into the technical detail and justification of the existence of the negative numbers. The construction of the negative numbers at this point would cloud our view and make us lose sight of the over all picture. We can view the existence of the negative numbers as guaranteed by construction. Thus the integers consist of the counting numbers, their negatives and zero. Ordering comes naturally with the counting numbers together with zero. Successor comes after each number, starting from zero. It is not clear that the ordering is naturally extended to the negative numbers. If we view negative numbers as deficits, that are similar to money owed or quantity owed, then a natural ordering would be the same as that of counting numbers with zero at the start. One would have 1 comes after 0, 2 comes after 1, 3 comes after 2 and so on. Then 0 would be in a unique position as the starting number for both the counting numbers and their negatives. This would be fine until we view these numbers as distances. A deficit distance would assume an ordering as 1 comes before 0 and 1 comes after 0, ..., basically dictated by direction. So for the ordering on the negative to be consistent we would have 1 comes after 2 and 2 comes after 3 and so on. The number zero does play a unique role in this ordering and an even greater part in fact is played by the counting numbers if we were to make meaning of ordering (in an algebraic context).
The Rational Numbers

A ratio of 1 : n gives a fraction 1 for each counting number n. Then addition of m of n this 1 means m . This representation is not unique we have to allow for cancellation as 1 n n 2
2 Ng Tze Beng
means the same as 2 . There is the question of deciding when two representations are the same 4 without having to know the common factor of m and n. We say a and c are the same if and d b only if a d = b c. We write a = c when this happens. We can define negative fractions as b d similar to how we define integers as the solution x + r = 0 for each fraction r. The collection of the fractions, their negatives and zero constitute the rational numbers Q. There is another way of constructing the rational numbers from the integers by set theoretic consideration, a more mathematical way. We are keeping to our theme of extending a system to include the solution of all linear equations. Ordering does not come easily this time with the rational numbers. The set of "positive" rational numbers has the following properties. It should of course contain the counting numbers. For any two "positive" rational numbers p and q, A. p + q is again a "positive" rational number and B. p q is again a "positive" rational number. Notice that the counting numbers satisfy these two properties. This meaning of "positive" is artificial and unnatural. For one thing we know it ought to apply to the counting numbers and it does as we shall see later. So at least it is consistent with our perception of what positive would mean. Property B involves multiplication. It is easily seen that multiplication of a counting number by 1 gives a negative integer and so multiplication of a fraction gives us a negative rational number. We now additionally insist that this set of "positive" rational numbers, together with its negative, that is the set consisting of the result of multiplying each "positive" rational numbers by -1, and zero form the entire set of the rational numbers. Can we do this? That is do we have a candidate for this set of "positive" rational numbers? Yes. Our construction of the rational numbers involves the following ingredients, the fractions, their negatives and zero. Plainly the negative fractions are the result of multiplication of the fractions by 1. Obviously the fractions satisfy Properties A and B. Is our insistence misplaced? Is there a set of different "positive" rational numbers than the fractions? The answer is no. Our insistence does pay off. This definition of "positive" would capture the essence of the meaning of positive. Indeed, it gives it new meaning. Note that 1 is "positive", a notion we would accept readily. But with the new meaning, it would require some thought. We would use a contradiction argument to show this. If 1 is not "positive", then it's negative 1 would be "positive" and by Property B, (1)(1) = 1 would be "positive", contradicting our assumption that 1 is not "positive". Then by Property A, all the counting numbers, being defined successively by adding 1, are "positive". Also note that for any counting number n, 1 is "positive". This can be verified as follows. If 1 is not n n "positive", then 1 is the multiplication of a "positive" number by 1 because it is not 0. Thus n 1 is "positive" and so since n is "positive", 1 = 1 $ n would be "positive" by Property B n n contradicting 1 is not "positive". Therefore, we conclude that 1 is "positive". It then follows n
3 Ng Tze Beng
from Property B that any fraction m is "positive" for any counting numbers m and n, since n m = m $ 1 . Thus our fractions are "positive". Then the "positive" rational numbers are n n precisely the fractions. This is because if there is a "positive" number p not a fraction, then since p ! 0, p is a fraction and so p is "positive" (since any fraction is "positive") and so p is not "positive" contradicting p is "positive". We call this subset of "positive" rational numbers, a positive cone. It is precisely the set of fractions. The positive cone or the fractions would serve as a kind of reference for the ordering. It is a natural division of the rational numbers into two parts, a special part that decides the "direction" of an ordering and another. There is always a division of the rational numbers at any point into the 'left' and 'right'; what we needed is a reference point, zero, and a translation operation to give meaning to 'left' and 'right'. For any two rational numbers a and b, we say a is greater than b ( a > b) if and only if a b is "positive", i.e., a b belongs to this special part. This ordering is consistent with the ordering on the counting numbers. This is seen as follows. For any counting number n, n + 1 > n because (n + 1) n = 1 is "positive". Since n +1 is the successor of n, this ordering is consistent with the previous ordering determined by the sequence of successor followed by successor. In particular for any fraction r, r > 0 because r 0 = r is "positive". If we now define any rational number r to be positive when r > 0, then positive would mean the same as "positive". In here is buried the notion of "reality", a reference to the continuous nature of time, past, present and future. We cannot in any way pinpoint what the next instant of time is, but we can say the future is what comes after the present instant. The next instant does not exist until it becomes the present instant. What lies beyond the present is the future. The difficulty in describing time is the same as the difficulty in describing reality.
What We Would Like the Real Numbers to Possess

If we can view the real numbers as the extension of the rational numbers, then we would want the properties that the rational numbers possessed that are so useful, to carry over to this extension. We shall describe in abstract terms these properties. The rational number system is an example of an object mathematician calls a field. It is a set F that comes with two binary operations called addition (+) and multiplication (), two
4 Ng Tze Beng
unique elements called respectively 0 and 1, two unary operations, one on F denoted by : F d F and the other on F {0}, denoted by & : F {0} d F {0}satisfying the following 9 properties. For all a in F, 1. a + 0 = 0 + a = a ; 2. a + ( a) = ( a) + a = 0. For all a, b and c in F, 3. a + (b + c) = (a + b) + c; 4. a + b = b + a; 5. a(b + c) = ab + ac. For all a in F {0}, 6. 1 a = a 1 = a 7. a (*a) = (*a) a = 1. For all a, b and c in F {0}, 8. a (b c) = (a b) c; 9. a b = b a.
(Associativity) (Commutativity) (Distributivity)
(Associativity) (Commutativity)
The unary operation * for the rational numbers Q corresponds to taking reciprocals on non zero rational numbers. The other operations are suggestive of the symbols. A totally ordered field F is a field F together with a positive cone P such that 0 does not belong to P , the union of P, its reflection P = { a : a belongs to P} and {0} is equal to F and P satisfies the following two properties that for all a and b in P, (A) a + b belongs to P and (B) a b belongs to P. The ordering, '>' , called a total ordering on F , is defined by b > a if and only if b a belongs to P. Thus for any x in F, x > 0 if and only if x belongs to P. The rational numbers Q satisfy all the 9 properties with the usual operations of addition, multiplication, taking negatives and reciprocals and has the total ordering described earlier with the (positive) fractions as the positive cone. To see the desirable property the real number system should possess, we may have to reinvent the whole system of representing numbers. In the theory of mensuration, there are numbers or quantities that are not commensurable. Take for instance 2 the square root of 2. Is this a number? Geometric intuition says it is. One thing is sure - we can "approximate" 2 by fractions. There are rational numbers as close to 2 as we like 'before' and 'after' 2 . We cannot pin down 2 as a rational number. It is not a symbol readily understood as 2 or 1 . But what we can say is this. If there is such a number, then its 2 square would give us the integer 2. We can be bold. We can extend, in any sense as we would, our rational numbers to some system containing the solution of the equation x2 = 2. But then we would just open up a Pandora's box. What about 3 , 5 , 7 , ..., p , p a prime and so on? It
5 Ng Tze Beng
is well-known none of these can be rational numbers. What about cube root of 2? These are solutions to polynomial equation of the form xn = p where p and n are counting numbers. What about solution to all polynomial equations? It then becomes an impossible task to describe all these "numbers". In particular there are "numbers" that are not the solution of a polynomial equation such as the Euler constant e and . So extending our rational numbers this way would not include these numbers. But what is plausible is that no matter what their origins may be, there are rational numbers as close to these numbers as we like 'before' and 'after' these numbers. This forms the basic concept of the cut of Dedekind. We have to think of numbers differently as if there is a hierarchy of numbers. We may not know what 2 is but we know there are rational numbers as close as we like on the 'left' of it or less than it if we can give an ordering on our set of real numbers. This gives enough information about 2 for all practical purposes. Indeed, it is a collection of rational numbers that can give us all the information we required about 2 . Hence we need a collection of rational numbers to describe a number. We shall come back to this later when we embark on making this more precise. We say the totally ordered field F has the Archimedean Property if for all x > 0 in F, and for all y > 0 in F, there exists a counting number n such that (n1) x > y. We can rewrite the last inequality as 1 % y < x or (&(n1)) % y < x. What this says is that given n1 any x and y > 0 in F, no matter how small x is, we can find a counting number n such that 1 % y < x if we identify 1 with 1 . Obviously the rational numbers Q has the Archimedean n n n1 Property. This is a property that we would wish the real numbers to have. We would need to add a new property that would tell us that the set of real numbers does exist in a different sense. We know there are rational numbers arbitrarily bigger than 2 . We can think of all the rational numbers below or less than 2 . In a way 2 would be the largest such number if it exists, bigger than all the rational numbers below 2 . The existence of such a number would have guaranteed the meaning of 2 . But of course we would have to think of rational numbers in a different way. To describe more precisely what we mean, we make the following definition. Consider a subset A of F. A is said to be bounded above if there exists x in F such that for all a in A, a [ x. Here a [ x means a < x or x = a. We say A is bounded below if there exists y in F such that for all a in A, y [ a. The number x is called an upper bound for A and y a lower bound for A. We say A is bounded if it is both bounded above and bounded below. If A is bounded above, then it has an upper bound. It is natural to ask if it has the smallest such upper bound. That means if M is the smallest such upper bound, then of course M is in F and for any x in F with x < M, x cannot be an upper bound for A and so consequently, there exists a a0 in A such that x < a0.
6 Ng Tze Beng
Definition 1. M is the least upper bound or supremum (sup) of a subset A of F if for all a in A, a M and for any x < M, there exists b in A such that x < b. A more descriptive way of describing M is this: For any number x less than M, we can always find an element b in A such that x < b M. Similarly we can define the greatest lower bound or infimum of A. Definition 2. m is the greatest lower bound or infimum (inf) of a subset A of F if for all a in A, m a and for any x > m, there exists b in A such that b < x. We can thus characterize m by saying that for any x > m, we can always find an element b in A such that x > b m The notion of supremum or infimum would be in vain if they do not exist. We would like them to be included in our consideration. A totally ordered field in which every bounded subset has an infimum and a supremum is special in that the boundaries of the bounded subsets are elements in the field. This prompts the next important definition. Definition 3. A totally ordered field F is complete if every non-empty bounded above subset of F has a supremum. The significance of this definition is that the supremum is a member of F. That means any bounded above subset has its 'upper' boundary residing in F and there is no room for a gap to exist in F. The term 'complete' has several meanings. The present meaning is sometimes referred to as order complete. We shall see later how the several meanings converge. This property is new. It is desirable for 2 to have a meaning. The rational numbers Q is not complete. Take for example the subset A = {x in Q: x > 0 and x2 <2 } of Q. It does not have a supremum (in Q). Note that we cannot as yet write 2 as its existence has not been established. We can only talk about it hypothetically. Even if we assume that its existence has been established, we still have to know how the usual ordering on the rational numbers will place 2 by some extension of the ordering. That 2 is not a rational number can easily be p shown as follows. Suppose 2 is rational, say 2 = q , where p and q are counting numbers with no common factors other than 1. Then p = 2 q and so p2 = 2 q2 . Thus p2 is even. Since product of odd numbers yields odd number p is even and so p = 2 k. Thus 2 q2 = p2 = 4 k2 and we get q2 = 2 k2 . A similar argument yields q is even. Hence 2 is a common factor of p and q contradicting that p and q have no common factors other than 1. Thus 2 will belong to a different scheme of things. A is plainly bounded above for we see that for any a in A a < 2. We shall now show that A has no supremum.
7 Ng Tze Beng
Suppose A has a supremum M. Then M a for all a in A and that if k < M, then there exists b in A such that k < b. By definition of A, a2 < 2 for all a in A. Also a2 M 2 . We can 2 compare M 2 and 2 to produce a contradiction. Suppose M 2 > 2. Then M 2 > 0 since M+2 M 2 2 = 2M + 2 . Then k < M. Note that plainly M +2 > 0. Let k = M M+2 M+2 4(M + 1) 2 2(M + 2) 2 2(M 2 2) k2 2 = = > 0 since M 2 > 2. Thus k 2 > 2. But since k < M, (M + 2) 2 (M + 2) 2 there exists b in A such that k < b. Thus k2 < b 2 < 2. This contradicts k 2 > 2. Therefore, M 2 2 . Since M 2 2 because M is a rational number, M 2 < 2. We shall now derive another 2 contradiction. So we have 2 M 2 > 0. Let now k = M + 2 M = 2M + 2 . Then k > M . Also M+2 M+2 2(M 2 2) 2 2 2 k 2= < 0 since M < 2. Therefore, k < 2. Hence k belongs to A and so k M . (M + 2) 2 This contradicts k > M. Consequently, these two contradictions imply that M = sup A does not exist. The following is a variation or equivalent definition for completeness. First, note that for any non-empty bounded below subset A of F, A = {a: a belongs to A} is bounded above. In particular, inf A = sup ( A). Thus if the supremum exists for any non empty bounded above subset of F, then the infimum too exists for any non-empty bounded below subset of F. Definition 3'. A totally ordered field F is complete if every non-empty bounded below subset of F has an infimum. It is clear from the above that Definition 3 implies Definition 3'. It can be similarly observed that for any bounded above subset A of F, A = {a: a belongs to A} is bounded below and sup A = inf ( A). This will supply the argument for proving that Definition 3' implies Definition 3. If we assume we know how to construct the real numbers, then the following tells us just what it is. Theorem 1. The real numbers R is a complete totally ordered field. There is essentially one such complete ordered field. This is not to say that there is exactly one such complete totally ordered field but that any two are isomorphic. We interpret this to mean that for all intent and purposes they are the same although they may be constructs of a different nature. Proposition 2. The real numbers R has the Archimedean Property. That is to say for any x, y > 0 in R, there is a counting number n such that n x > y ---------- (*). (Here we are using the notation inherited from the rational numbers.)
8 Ng Tze Beng
When a totally ordered field has the Archimedan Property, we say it is archimedean. Thus R is archimedean. Proof of Proposition 2. We shall prove Proposition 2 by contradiction. Suppose R is not archimedean. Then by negating the statement (*), we get there exists x, y > 0 such that for all counting number n, n x y. ------ (**) Take the set K = { n x : n a counting number}. Then by (**) K is bounded above by y and is non-empty. Because R is complete, the supremum M of K exists. Hence for any counting number n, n x M since n x belongs to K. Now (n + 1)x belongs to K too. Therefore, (n + 1)x M. Thus for any counting number n, n x M x < M. ------------------(***)
Thus M x is an upper bound for K. Because M x < M and that M is the supremum of K, there is an element n0 x in K, for some counting number n0, such that M x < n0 x. But by (***) n0 x M x. This contradicts M x < n0 x. Therefore, R is archimedean. Real numbers are hard to think of conceptually. When we say take a small number > 0, we like to think of as a rational number since we are more comfortable with the rational numbers. For all practical purposes, this is what we need to think of . We may indeed just say take a small rational number > 0 instead. The following justifies this. Corollary 3. For any > 0, there is a counting number n such that 1 < . n Proof. By the Archimedean Property of R, there exists a counting number n such that n > 1. Therefore, 1 < . n Note that 1 is a rational number. So the Corollary says that for any > 0, no matter how n small is, we can find a rational number 1 such that 0 < 1 < . So for all pratical purposes, in n n 1 instead. place of , we can use n Now that we have recognised a class of numbers which consists of numbers that are not rational numbers and that there is an ordering that applies to the whole of R, we might ask ourselves the question, "How often can we find a rational number or for that matter, irrational number ?" Very often is the answer. In mathematical terms we mean the rational numbers or the irrational numbers are dense. The following Corollary gives meaning to the term 'density of the rational numbers'.
9 Ng Tze Beng
Corollary 4. For any x and y in R and x < y, there exists an integer n and a counting n number m such that x < m < y . A descriptive way of stating Corollary 4 will be "between any two real numbers there is a rational number". Proof of Corollary 4. The proof goes like this. Take two real numbers x and y such that x < y. Then y x > 0. It follows by Corollary 3 that there is a counting number m such 1 that m < y x. The rest of the proof will be divided into 3 cases. The easiest case will 1 1 be when x = 0. Then we have 0 < m < y and so the required rational number is m for this case. The second case is when x > 0. We now invoke the Archimedean Property of 1 R. By this property there is a counting number n, such that n m > x. Having established the existence of such an integer n, we can then by successively taking one away from this 1 1 number n to obtain the least integer n such that n m > x. That means (n 1) m [ x. Therefore,
1 1 n y = ( y x) + x > m + (n 1) m = m > x. n For this case, the required rational number is m . The remaining case is when x < 0. That is x > 0. Then by the Archimedean property there is a counting number n such 1 1 that n m > x. As before we choose the least integer n such that n m m x. Then we have 1 (n 1) m < x. Therefore, 1 n 1 y = ( y x) + x > m m = (1 n) m > x. For this case the required rational number is 1 m n . This completes the proof.
Having established the density of the rational numbers, we expect that the irrational numbers are also dense in R. More is true here. The irrational numbers are more numerous than the rational numbers. This statement will make sense only when we have some means of "measuring" subsets of the real numbers. Indeed, the "measure" of the set of rational numbers is zero but not so for the set of irrational numbers. We would need a theory of measure to establish this. The fact that R is uncountable and the rational numbers Q is countable gives us some idea of the difference in "size" of the set of irrational numbers and the set of rational numbers. A set is said to be countable if we can match its elements one to one with elements of the counting numbers. It can be shown that Q is countable though not finite. But it is much harder to show that R is not countable. One can do this by showing that the real numbers between 0 and 1 is not countable. This can be shown by way of contradiction. First, by assuming that we have a matching function from the counting numbers to the real numbers between 0 and 1 and thus we can write them as a sequence. Then by assuming that each term of this sequence can be written as an infinite decimal and using this sequence of infinite decimals to produce a number different from any term of this sequence and thus showing that we can never
10 Ng Tze Beng
have a matching function. This approach will need some criterion to distinguish infinite decimals converging to different limits. For now we are content with the following. Corollary 5. For any two rational numbers a and b with a < b, there is an irrational number such that a < < b. Proof. The proof of this Corollary is by actually producing the required irrational number by making use of a known irrational number. Take an irrational number k > 0. For instance we can take k = 2 . By the Archimedean Property of R, there is a counting number n such that n (b a) > k. That is, k a + n < b. k k Then a < a + n < b. Since k is irrational, a + n is also irrational because a is rational k k and n is irrational. Take = a + n to be our required irrational number. This establishes the truth of this corollary. Now here is a curious observation. Each integer is separated from the nearest integer by one. We can expect to find an integer between x and x +1 for any real number x. This is stated more precisely as follows. Corollary 6. For any real number x, there is an integer n0 such that x < n0 x + 1. Proof. If x is an integer, then we only have to take n0 to be x + 1. Now assume x is not an integer. Suppose x > 0. Then by the Archimedean Property of R, there is a counting number n such that n = n1 > x. Then take the least such integer N with N > x. Thus we have N 1 x. Therefore, we have x < N x + 1. So take n0 = N. Observe that we actually have x < N < x + 1, since x + 1 is not an integer. Now for the case x is not an integer and x < 0. Then as before there exists an integer M such that x < M < x +1. Thus x 1 < M < x and so x < M +1 < x + 1. For this case take n0 = 1 M . This completes the proof. In the next few chapters we shall construct the real numbers and show that it is a complete totally ordered field.
11 Ng Tze Beng
CHAPTER TWO
A Cut, An Instant
Because I know that time is always time And place is always and only place And what is actual is actual only for one time And only for one place .............. T.S. Eliot
To begin the construction of the real numbers, we shall take another look at the object we shall use, namely the rational numbers in a different way. As we stand at the threshold of time dividing the past and the future, there is an urge to move and yet not moved for time's ceaseless motion like an invisible hand divides the past and future at any instant now and forever. Pondering upon the total ordering on the rational numbers, we have a situation that mirrors the ever present time. At any point P (either to be made precisely or a rational number), there is according to the total ordering a division of the rational numbers into two classes of numbers, the class whose numbers are all greater than P mimicking the future of P and the class whose numbers are all less than P mimicking the past of P, a cut in time. The nature of the point P or the cut is the subject to be discussed. This cut in time is infallibly associated with an instant, the present instant. The untethered instant has order amidst the chaos. We cannot tell when it is and yet we know. The set of rational numbers then has a new meaning and in this new meaning is born the seed of the real numbers.
A Cut in Time
We cannot quite yet describe the present but we can for practical reason describe what comes close to being the present. A measure of time, the second, the humanly imaginable unit of time is our yardstick. As we speak, the present becomes in a fraction of the second the past and the future seems in a fraction of the second like the present. Time never stands still. We can in our humanly possible way describe what comes close to an instant as a cut in time. Definition 1. A cut of Q is an ordered pair of subsets of the rational numbers (L, R) satisfying the following three properties: (1) L , R . (2) L R = , L R = Q. (3) If x is in L and y is in R, then x < y. The subset L is called the left set of the cut and R the right set of the cut. By Property (2), every cut is determined by its left and right sets, each of which determines the other. Therefore, we shall identify a cut with its left set.
Chapter 2 A Cut, An Instant
Thus each 'point' or each instant is given by a cut. Each 'point' as time has its past and future. If we are given a reference point, then we can begin to think of each rational number as a cut. This then gives new meaning to the rational numbers. Definition 2. A subset of Q is called a real number if it satisfies the following two properties. (A) is the left set of some cut; and (B) has no greatest element with respect to the total ordering on Q.
L
Time Line
R P
The present instant, the fleeting moment, once you think you have it in your grasp, it has already escaped you; yet we know the present instant. A point, an instant has taken on a new meaning as a cut in time. A 'real' number is not a number yet but part of a continuum of time in analogy. The abstract notions of addition and multiplication have yet to be invented. Property B models the fleeting nature of an instant. Technically not all cuts are real numbers. Take for example the following cuts. (1) (L, R) = { x Q: x2 < 2} { x Q: x2 > 2}. (2) (L, R) = { x Q: x < 2} { x Q: x 2}. (3) (L, R) = { x Q: x 2} { x Q: x > 2}. Cut (1) and (2) are real numbers but cut (3) is not a real number. The cuts (2) and (3) correspond in some sense the integer 2, but we do not want both. We identify a real number with the left cut that has no greatest element. Like the essence of an instant, knowing but unattainable, the real number is represented by the left cut that has no greatest element. Out of these special partitions of rational numbers, identified as the left hand sets of the cuts and hence out of the basic building blocks in set theory, we shall realise the real numbers. We shall have to redefine what seems so natural in the rational numbers. The identification of the cuts with their left hand sets provides us with the ease to redefine operations of addition and multiplication and others in terms of the basic operations in set theory. The following is a way of deciding when a subset of the rational numbers is a real number under our identification. Lemma 3. A subset of Q is a real number if and only if the following four conditions are satisfied. 1. . 2. Q. 3. For any x in , if y is a rational number such that y < x, then y is also in . 4. has no greatest number.
13 Ng Tze Beng
Proof. Let L = and R = Q . Then conditions (1) and (2) imply that L and R . For any x in L and any y in R, x < y. This is because if on the contrary that x y, then by condition (3) above y is in . But since y is in R = Q , y is not in . This contradicts that y is in . Therefore, by Definition 1, (L, R) is a cut. Condition (4) then says that L has no greatest number. Therefore, is a real number. Conversely if is a real number, then the conditions (1) to (4) are automatically satisfied. The essence of a cut is captured in the following lemma. It says that there are points on the left and right of a cut that are as close to one another as one wishes. Lemma 4. 1. If is a real number and n is a counting number greater or equal to 1, then there exists (rational numbers) a in and b not in such that ba < 1. n
2. For any rational number a in a real number , there exists a counting 1 number m such that a + m c . Proof. Part 1. The real line will help us to visualise this lemma. Look at the following picture carefully.
a
< 1/n
Since is a real number, we can take an integer c not in and an integer d in . Then starting from c we shall add lengths of 12 to give rational numbers that are n 1 apart. Define x = d + i . Then x = d, x , x , x , ... , x 2 consecutively 2 0 1 2 3 i n (cd) = c n2 n is a sequence of rational numbers such that x0 = d < x1 < x2 < x3 < ... < x n 2 (cd) = c. Therefore, for some integer j such that 0 j < n2(c - d), xj is in but xj+1 is not in . Take b = xj+1 and a = xj . Then for n > 1, b a = 12 < 1 . For n =1, use n + 1 instead n n 1 < 1. as above to obtain b and a such that b a < n+1 n Proof of Part 2. Take any rational number a in the real number . Suppose on the 1 contrary that we cannot find such a number m with a + m c . Then for all counting number n, a + 1 . That is, for all counting number n, x < a + 1 for all x in . n n This implies that for each x in , x a. This is shown by way of contradiction.
14 Ng Tze Beng
Suppose on the contrary that x > a. Then by the Archimedean property of Q, there exists a counting number p such that p (x a) > 1. That is 1 < x a. Hence p 1 . This contradicts x < a + 1 . Therefore, for each x in , x a. Since a is in x>a+ p p , a is a maximum element in , contradicting that does not have a maximum element because it is a real number. Thus for each a in , there exists a counting 1 number m such that a + m c . A cut has a natural ordering arising out of a subset inclusion. Like the marching of time extending into the future, this analogy aptly describes the ordering. We make the definition below. Definition 5. For any two real numbers and , we say if and only if .
Let the set of real numbers be denoted by R. Clearly this ordering ' ' is a reflexive, transitive and antisymmetric relation on the set R. What does this mean? ' ' is reflexive means that for all in R, . It is obviously true since for any set . Transitivity means if and , then . Clearly if and , then and and so which means that . ' ' is antisymmetric means that if and , then = . This is plainly true since and imply that = . Lemma 6. This ordering ' ' on R is a total ordering. Proof. We have shown above that ' ' is a partial ordering on R. We need to show that any two real numbers and are comparable. That is either or . Since both and are subsets of Q, either or M . If , then and we have nothing to prove. It remains to show that if M , then . Now if M , then there exists a rational number x in such that x is not in (x ) . Then for any y in , y < x by Property 3 of Definition 1 because is a cut. Therefore, by Property 3 of Lemma 3, for any y in , y is in since x is in . Thus . Therefore, .
15 Ng Tze Beng
Then R is complete in the sense of order. We shall show this below. First, let us examine what we have not done so far. We have not yet define the operations of addition and multiplication on R. The only structure we have on R at the moment is a total ordering. Lemma 7. Let S be a subset of R which is bounded above. Then S has a least upper bound or supremum in R. Set theoretically the supremum is easily found. But we do need to show that we can actually obtain a real number this way. Much of the proof of Lemma 7 goes in showing this. Remember each element in S is a subset of the rational numbers. Proof of Lemma 7. Define = 4 { : c S }.Then for all in S. Remember that each in S is a cut and so is a subset of the rational numbers. Therefore, is a subset of the rational numbers Q. We need to show that is a real number and that it is the supremum of S. We are given that S is bounded above. Therefore, there exists a real number such that for all in S. That means for all in S. Therefore, = 4 { : c S} ` . If we can show that is a real number, then . We shall now use Lemma 3 to show that is a real number. Since for each in S, we have then that . Also Q because and Q. Take now any x in . Then x for some in S. Thus for any y in Q with y < x, y is in by Property 3 of Lemma 3 since is a real number. It now remains to show property 4 of Lemma 3 for ; that is has no maximum element. We shall show this by contradiction. Suppose on the contrary that has a maximum element l. Then l is in some in S. Also for all y , y l. This is also true of all y in this particular , since . Thus l would be the greatest element in . This contradicts that has no greatest element since it is a real number. This completes the proof that is a real number. Since for any in S, 4 { : c S} = , for any in S. Therefore, is an upper bound for S. Now let be any other upper bound of S. Then for all in S. That means for all in S. Therefore, = 4 { : c S} ` . Hence . This shows that is the least upper bound or the supremum of S. We have thus proved the following. Theorem 10. R has a complete total ordering. The construction of the real numbers makes it stand apart from the rational numbers. Firstly, it is constructed out of subsets of the rational numbers. For second, the set of rational numbers is definitely not a subset of it, at least not in a natural sense. To make sense of the rational numbers in this new as yet to be defined system of real numbers, we shall embed our rational numbers into R in such a way that it is compatible with the ordering on the rational numbers.
16 Ng Tze Beng
The Embedding of the Rational Numbers in The Real Numbers

Define an embedding : Q R by (a) = { x Q: x < a} for any rational number a. This is the only natural way of embedding Q and gives the new meaning of Q that was mentioned earlier. This is well defined for (a) is a real number. Why? Obviously (a) is a cut that does not have a maximum element and so by definition (a) is a real number. The image of Q under is truly a copy of Q. Until addition and multiplication are defined on the set of real numbers we cannot show that the image behaves just like the rational number system. This will be done in the later chapter. We now show that is injective and is compatible with the ordering on the rational numbers. Mathematically this is summarised as follows. Lemma 11. : Q R is injective. Furthermore, for any rational numbers a and b a b (a) (b) . Proof. We shall show that is injective. More precisely, we shall show that whenever (a) = (b), then a = b. (a) = (b) implies that {x Q: x < a} = {x Q: x < b}. If a < b, then a {x Q: x < b}. Take c = a + b . We have then a < c < b, c 2 {x Q: x < b} and c {x Q: x < a}. Therefore, {x Q: x < b} {x Q: x < a} contradicting {x Q: x < b} = {x Q: x < a}. Thus a must be greater or equal to b. We can show similarly that a cannot be greater than b. Thus a = b. Hence is injective. If a b, then (a) = { x Q: x < a} { x Q: x < b} = (b) . Therefore, (a) (b). Conversely, if (a) (b), then { x Q: x < a} { x Q: x < b}. It follows that a b, for otherwise a > b would imply that { x Q: x < a} M { x Q: x < b} since we would have c = a + b { x Q: x < a} and c { x Q: x < b}. This 2 completes the proof of Lemma 11. Thus we now have a new model of the rational numbers embedded in the real numbers, We have yet to define multiplication and addition on R and yet to show that the embedding respects addition and multiplication. We shall come back to this in the later chapter. Having known that the embedding respects ordering, Lemma 4 will then has new interpretation as given below. Corollary 12. If is a real number, then for any counting number n, there exist rational numbers a and b such that (a) < (b) and (b a) < ( 1 ), n where the strict ordering ' < ' is defined by < if and only if and .
17 Ng Tze Beng
Proof. By Lemma 4, for any counting number n, there exist rational numbers a in and b such that b a < 1 . Since a , by Property 3 of Lemma 3, any rational n numbers x < a belongs to . Therefore, { x Q: x < a} . That means (a) , therefore (a) . ecause a (a) and a , (a) . Thus (a) < . Since b , for any x in , x < b because b is in the right set of the cut . Hence { x Q: x < b}= (b). Therefore, (b). By Lemma 11, (b a) ( 1 ). ecause n b a < 1 , b a c ( 1 ). Now that we also have b a (b a), (b a) ( 1 ). n n n 1 ). This completes the proof. Therefore, (b a) < ( n Corollary 12 says that for any real number , there are rational numbers (the embedded kind) before and after that are arbitrary close to one another. In the next chapter we shall define addition on R.
18 Ng Tze Beng
CHAPTER THREE
Relearning Addition
Like a child who has forgotten how to walk Like a bird who has forgotten how to fly We shall learn, step by step, Carefully, gingerly, how to add Not by natural designs, but by constructs That spring from our forgotten past.
Addition on the real numbers comes naturally but subtly as we shall observe in our definition. Definition of Addition. For any and in R, define + = { x + y : x , y }. This looks natural but we need to know if it gives us a real number. It would not make any sense if it is not a real number. Note that + . We shall show that + Q. Since is a real number, there exists a rational number c such that for all x in , x < c. Likewise there exists a rational number d such that for all y in , y < d. Then for any x in and any y in , x + y < c + d. Therefore, c + d + . Hence + Q. Now take any x in and any y in . For any rational number z < x + y, z x < y . Hence z x by Property 3 of Lemma 3 Chapter 2 since is a real number. Then z = x + (z x) + . ence + satisfies Property 3 of Lemma 3 Chapter 2. It remains to show that + has no maximum element. We show this by way of contradiction. Suppose + has a maximum element k in + . That is, for all r in + , r k and there exists x in and y in with k = x + y. Therefore, r k = x + y. Then r x y for any r in + . Now take any l in , x + l + . Thus by the above argument, for any l in , ( x + l ) x = l y (taking r = x + l ). Hence since y is in , y is the maximum element of , contradicting that has no maximum element (because is a real number). This contradiction tells us that + has no maximum element. We can thus conclude by Lemma 3 of Chapter 2 that + is a real number. We have thus proved the following. Lemma 1. For any and in R, + is a real number.
Chapter 3 Relearning Addition
Hence the usual properties of addition will be rules to be verified. They are not obvious. We need to check these rules by different means than we are accustomed to. Lemma 2. Let , and be any real numbers in R and a, b any rational numbers in Q. Then the following holds. 1. + = + . 2. ( + ) + = + ( + ). 3. + (0) = . 4. + + . 5. (a + b) = (a) + (b). Before we embark on a proof of this lemma, let us examine what it says. Property 1 and Property 2 are the usual properties one expects from any definition of addition. Property 3 says that (0) is the zero for this addition. Notice the difference from the old zero. It is a subset of the rational numbers. Property 4 says that addition respects the ordering on the real numbers R. Finally Property 5 says that the embedding of the rational numbers Q into R respects both addition on Q as well as on R. Proof of Properties 1 and 2. Obviously + = { x + y : x , y } = { y + x : x , y } = + since x + y = y + x for any rational numbers x and y. Now ( + ) + = { k + z : k + , z } = { (x + y ) + z : x , y , z } = { x + ( y + z) : x , y , z } since for any rational numbers x, y and z, (x + y ) + z = x + ( y + z) = { x + l : x , l + } = + ( + ). . Proof of Property 3. Recall that (0) = {x Q: x < 0}. Then + (0) = { x + y : x , y (0) }. Since for any y in (0), y < 0, therefore for any x in , x + y < x. Since is a real number and x is in , by Property 3 of Lemma 3 Chapter 2, x + y is in . Hence for any x in and any y in (0), x + y is in . Therefore, + (0) . Take any a in . By Lemma 4 Part 2 of Chapter 2, there exists a counting number n such that a + 1 c . n hen a= a + 1 + 1 c + (0) n n
20 Ng Tze Beng
(Commutativity) (Associativity)
since 1 is in (0) because 1 < 0 . Thus + (0). Hence + (0). n n Therefore, = + (0). Proof of Property 4. if and only if . his means any element x in is also in . herefore, + = { x + y : x , y } { x + y : x , y } = + . Hence + + . Proof of Property 5. Let us write down the three subsets that are involved here.
(a) = {x Q : x < a}; (b) = {x Q : x < b} and (a + b) = {x Q : x < a + b}. Then (a) + (b) = { x + y Q : x (a) , y (b)} = { x + y Q : x < a , y < b} { z Q : z < a + b}= (a + b). Therefore, (a) + (b) (a + b). Next we shall show that (a + b) (a) + (b). Take any z in (a + b). Then z < a + b and so z a < b. Hence z a (b). By 1 Lemma 4 Part 2 of Chapter 2, there exists a counting number m such that z a + m 1 1 1 (b) . Therefore, z = a m + z a + m (a) + (b) since a m < a. This is true for every z in (a + b) and so (a + b) (a) + (b). Hence (a + b) = (a) + (b). Lemma 2 Property 3 says that the zero for the real number is (0). Notice that the rational number 0 is not in (0). We also need to show that (0) has the property that the rational number 0 enjoys, namely that each element in R has an additive inverse such that their sum gives (0). We want to define a unary operation R R that assigns each in R its additive inverse . Definition 2. For any in R, define = { x Q : x < y for some y }. There are two things we need to do. We need to show that is a real number and that indeed + () = (0). Lemma 3. For any real number , is a real number.
Before we embark on the proof of this lemma. Let us examine the rational behind the definition. The concept of additive inverse for a rational number is an extension of the definition of negative integers conceptually as discussed in Chapter 1. Both in concept and construction the additive inverse for an integer is strongly associated with a direction and the notion of distance. Envisaged in another way the additive inverse of an integer is the
21 Ng Tze Beng
reflection of the integer about the point 0. Reflection is multiplication by 1. It has the effect of reversing the ordering of any two numbers. Each real number carries with it a whole baggage of the rational numbers but without the structures. The additive inverse seems so unnatural to define. For distance, reflection is easily understood. What about a cut? A real number is represented by the left hand cut of a special cut. If we apply reflection to a CUT, the left hand set becomes the right hand set and the right hand set becomes the left hand set. Suppose is the left hand set of the cut (L, R). Then the reflection of this cut gives a cut ( R, L), where R is the set { r : r R} and L = { x : x L }.
=L (0) -L R
-R
This may not give you a real number for R may have the greatest number. To obtain a real number we must exclude the greatest number in R whenever it exists. There is a way to do this. Take any rational number y in R and consider its negative y in R. Then take all rational numbers x such that x < y. In this way if y is the greatest element in R, then the set {x Q ; x < y} will not have the greatest element. This explains the definition: = { x Q : x < y for some y }. From the point of view of a cut, it is easy to see that is the left hand set of some cut, which has no maximal element and so it is a real number. All the same we shall check if indeed it is a real number. Proof of Lemma 3. First of all . This is deduced as follows. Since Q , we can choose a rational number y . Take any rational number a > y. Then a since a < y. Thus . Let K = { y : y Q and y }. Then K Q because { y Q : y } Q since . Then K. This is because for any x in , there exists y , such that x < y. Therefore, x > y and so x and that means x belongs to K. This is true for all x in and so gives us the stated inclusion. Therefore, Q. Now take any x in . Then for any rational number b < x , b is in . This is because there exists a rational number y such that x < y and so b < y and that means b belongs to . Finally we shall show that has no maximal element. Suppose it has a maximal element M. That is for all x in , x M. Since M is in , there exists a rational number y0 not in such that M < y0 . M y0 Then by taking z = , we see that there is a rational number z such that M < z < 2 y0 . Obviously z and so z M . This contradicts M < z and so has no maximal element. Therefore, we have shown that satisfies Properties 1 to 4 of Lemma 3 of Chapter 2 and so is a real number.
22 Ng Tze Beng
Next we shall show that is indeed the additive inverse of , i.e., that it lives upto the meaning that the minus sign in front of it would suggest. We shall also show that the embedding of the rational numbers into the real numbers takes the additive inverse of a rational number, a in Q, to the inverse of its image, (a), in R. Lemma 4. 1. If is a real number, then + () = (0). 2. For any rational number a, ( a) = (a). 3. For any real numbers, and , ( + ) = ( ) + () Proof. Part 1. First we shall show that + () (0). Recall that + () = { x + y : x , y }. For any y in , there exists a rational number y0 not in such that y < y0 . Now for any x in , x < y0 since y0 is not in . Therefore, x + y < y0 + ( y0) = 0. Hence x + y belongs to (0). This is true for any x + y in + () and so + () (0) and + () (0). Next we shall show that any negative rational number is in + (). First, we make the following observation. By Lemma 4 of Chapter 2, for any counting number n, there exist rational numbers b and c such that 0 < b c < 1 . Therefore, (2b c) c < 1 and so 1 < c + (c 2b). Then c 2b is n n 2n in because c 2b < b. Hence c + (c 2b) + (). Therefore, for any counting number n, 1 belongs to + (). For any rational number q < 0, q > n 0. So by the Archimedean Property of the rational numbers, there exists a counting 1 1 number n0 , such that n0( q) >1. Therefore, q > n 0 and that means q < n 0 . 1 Since, we have just proved that n 0 belongs to + () and because + () is a real number, q belongs to + (). Since this is true for any rational number q < 0, (0) + () and so (0) + (). Hence + () = (0). Part 2. (0) = (a + ( a)) = (a) + ( a) -------------------- (*) by Lemma 2 (5). Thus adding (a) to both sides of the equation (*) gives (a) + (0) = (a) + ((a) + ( a)) Therefore, (a)= ( (a) + (a)) + ( a) by the associativity of addition (Lemma 2 (2)) and Lemma 2 (3). Hence (a)= (0) + ( a) = ( a). This completes the proof. Part 3. (( ) + ( )) + ( + ) = ((( ) + ( )) + ) + by the associativity of addition (Lemma 2 part 2) = ( + (( ) + ( ))) + by the commutativity of addition (Lemma 2 part 1) = (( + ( )) + ( )) + by the associativity of addition (Lemma 2 part 2)
23 Ng Tze Beng
= ((0) + ( )) + = ( ) + by Lemma 2 part 1 and part 3. = (0). Therefore, ( + ) = (0) + ( ( + )) by Lemma 2 part 1 and part 3 = ((( ) + ( )) + ( + )) + ( ( + )) = (( ) + ( )) + (( + ) + ( ( + ))) by Lemma 2 part 2 = (( ) + ( )) + (0) = ( ) + ( ) by Lemma 2 part 3.
Lemma 2, parts (1), (2), (3) and Lemma 4 says that the set of real numbers with the operation of addition is an abelian group with identity given by (0). Hence we have proved the following. Theorem 5. R with the operation of addition is an abelian group with (0) as its identity element. The embedding gives then a copy of the rational numbers in R, additively. This is summed up in the following theorem. Theorem 6. The embedding : Q R of Chapter 2 is a group monomorphism of (Q, +) into (R, +). Proof. This theorem is a consequence of Lemma 2 Part 5, Lemma 4 Part 2 and Theorem 5. We have now succeeded in copying the additive structure of the rational numbers onto a subgroup of R, namely the image of the embedding . The next task is to see how the multiplicative structure can be defined on R and how our embedding will also copy the multiplicative structure of the rational numbers Q onto (Q). The task seems complicated and we shall do this in stages in the next chapter.
24 Ng Tze Beng
CHAPTER FOUR
Rethinking Multiplication
Multiplication for the rational numbers is no longer an obvious generalisation of addition. Its rule has become embedded in the definition of the rational numbers. For the real numbers, it seems we have to learn the meaning of multiplication by using the ubiquitous multiplication on the rational numbers. What is even harder is to recover the old meaning of multiplication this way. The only thing we can say about the definition below is that it seems to be the only thing we can try. We shall define multiplication in stages, first on real numbers (0) and then extend to the whole of the real numbers. We shall begin by considering the subset of the rational numbers constructed out of two real numbers (0) described in the lemma below. This is going to be the subset that gives the multiplication of these two real numbers. Of course we shall need it to be a real number. The following lemma confirms it to be the case. Lemma 1. For any two real numbers (0) and (0), define the subset of the rational numbers L by L = (0) { x Q: x = ab for some rational number a 0 in and some rational number b 0 in }. Then L is a real number. The subset L seems unnatural as a candidate for the multiplication of and . But the "obvious" subset { x Q: x = ab for some a in and some b in } cannot be a candidate for the multiplication simply because it can be too large or bounded below by 0. Take for example (0) and (1). Then we have that the subset defined by = { x Q: x = ab for some a in (0) and some b in (1)} is the same as the set of rational numbers Q. This is seen as follows. Obviously 0 (1) implies that 0 . For any x > 0, x (0). Since 1 (1), for any x > 0, x = ( x)(1) is in . Also for any x < 0, 2x (0) and so since 1/2 (1), x = (2x)(1/2) is in . Thus = Q and it cannot be a cut and so not a real number. The additional condition in the definition of L is to ensure we obtain a real number again. Before we embark on the proof of Lemma 1, we state the following characterisation of when a real number satisfies > (0). We write > (0) if and only if (0) and (0). Lemma 2. For any real number , > (0) if and only if 0 . In particular, if > (0), then there exists a rational number a > 0 in . Proof. > (0) implies that (0) is a proper subset of . This means there exists a rational number x in such that x (0). This means that x 0. Therefore, 0
Chapter 4 Rethinking Multiplication
since is a real number. Conversely if 0 , then for any rational number q < 0, q again since is a real number. This means that (0) and so (0) . By Chapter 2 Lemma 4 part 2, there exists a counting number n such that 0 + 1/n = 1/n . Take a = 1/n. Then a > 0 and a (0). Therefore, (0) and so > (0). This completes the proof. Proof of Lemma 1. Obviously L since (0) L . Next we shall show that L Q. For any , (0), either or since the ordering is a total ordering. Without loss of generality we assume that , i.e., . Since is a real number, there exists a rational number x0 not in such that for all x in , x < x0. Since (0), x0 0 for if on the contrary x0 < 0, that is x0 (0), then x0 would be in , since (0) . Similarly, there exists a rational number y0 not in such that for all y in , y < y0. Then for any rational number a 0 in and any rational number b 0 in , we have that a < x0 and b < y0 and a b <x0 y0. It follows then x0 y0 {x Q: x = ab for some rational number a 0 in and some rational number b 0 in }; and so since x0 y0 0, x0 y0 L . Hence L Q. Now take any rational number z in L . We shall next show that for any rational number x < z, x L . Now z is in L implies either z (0) or z { x Q: x = ab for some rational number a 0 in and some rational number b 0 in }. If z (0) or z = 0 (that is z 0), then we have nothing to prove since any x < z implies that x (0) L . If z > 0, then z {x Q: x = ab for some rational number a 0 in and some rational number b 0 in } and there exist rational number a0 > 0 in and rational number b0 > 0 in such that z = a0 b0 . Now take any x < z , if x < 0, then x (0) and so x L . We are thus left with showing that for any rational number x such that 0 x < z, x is in L . For any of these rational numbers x, we have 0 x < a0 b0. Therefore, x/a0 < b0 and so x/a0 since is a real number and b0 is in . Hence x = a0 (x/a0) with a0 in and x/a0 in belongs to L . Next we shall show that L has no maximal element. We note the following trivial case. If any one of the real numbers or is equal to (0), then L is equal to (0) because the set { x Q: x = ab for some rational number a 0 in and some rational number b 0 in } is empty. Because L is equal to (0) and (0) has no maximal element, we have nothing to prove for this trivial case. We may now assume , > (0). Suppose next that L has a maximal element M. We shall derive a contradiction. By Lemma 2, since , > (0), there exist rational numbers a > 0 in and b > 0 in . Therefore, a b > 0 and is an element of L . Because M is the maximal number of L , M a b and so M > 0. Consequently, M (0). It follows that M { x Q: x = ab for some rational number a 0 in and some rational number b 0 in }. Hence there exist rational number a0 > 0 in and rational number b0 > 0 in such that M = a0 b0 . Then consider a0 = M / b0. Because a0 0, for any a in such that a 0, a a0. Now for any a in such that a > 0, a b0 M so that a M / b0 = a0 . Thus for any rational number a in , a a0 . That means a0 is a maximal number for contradicting that is a real number. Therefore, the maximal number M for L does not exist. Hence it follows from Chapter 2 Lemma 3 that L is a real number.
26 Ng Tze Beng
Definition 3. For any real numbers and (0), we define the product by = (0) { x Q: x = ab for some rational number a 0 in and some rational number b 0 in }. By Lemma 1, the product is a real number. It remains to check if this product defines a multiplication. There are some immediate consequences of the definition. Firstly (0) since (0) . Secondly if both and > (0) , then > (0) since 0 because 0 belongs to both and . Proposition 4. Let and be any real numbers (0). 1. = . (Commutativity)
2. If any one of or is equal to (0), then = (0). Proof. Simply look at the definition of the product.
= (0) { x Q: x = ab for some rational number a 0 in and some rational number b 0 in } = (0) { x Q: x = ba for some rational number b 0 in and some rational number a 0 in } = . Part 2 is a consequence of the fact that if any one of or = (0), then {x Q: x = ab for some rational number a 0 in and some rational number b 0 in }= . Definition 3 is the first of a series of definition to define the multiplication on the real numbers R. Before we extend the definition to the whole of R, the next thing we shall do is to define the multiplicative inverse of a real number > (0). Later we shall extend this definition of the inverse to all of R {(0)}. Consider the next construct obtained from any real number > (0). Define M = { x Q: x < 1/y for some y }. Let us examine this set. First of all, the inequality in the definition makes sense because for y , y > 0 since by Lemma 2, 0 and is a (left) cut. Since for any y , 1/y > 0, this set is really all rational numbers less than or equal to 0 together with the reciprocals of all rational numbers not in with the exception that if the complement of has a least number (as in (1)), we do not get the reciprocal of this number. This set looks naturally like the multiplicative inverse of . The usual multiplicative inverse of a non zero rational number is its reciprocal. Here is a set, so we shall have to take for its multiplicative inverse a collection of the reciprocals cleverly and to include at least sufficient rational numbers to give us a real number. Then the next thing to do is to show that it really is the multiplicative inverse.
27 Ng Tze Beng
Proposition 5. If > (0), then the subset M of the rational number is a real number. Proof. 1. M is obviously not empty since 0 M. 2. M Q. This is seen as follows. Since > (0), by Lemma 2 there exists a rational number x0 > 0 in . Then for any rational y , because is a real number, y > x0 > 0. Thus 1/x0 > 1/y for any y . For any x in M, there exists some y such that x < 1/y. Therefore, for any x in M, x <1/x0. Thus 1/x0 M and so M Q. 3. Take any rational number z in M. Then there exists some y such that z < 1/y. Hence for any rational number x < z, x < 1/y and so x M. 4. M does not have a maximal number. Suppose on the contrary M has a maximal number M. Then M x for all x in M. Note that M M implies that there exists some y such that M < 1/y. Take the rational number q = (M + 1/ y)/2 satisfying M < q < 1/y. Then q M and so M q contradicting M < q. Therefore, M does not exist. We are now ready to make our definition. Definition 6. For any real number > (0), define 1 = {x c Q : x < 1 for some y " }. y By Proposition 5 1 is a real number. An immediate consequence of the definition is that 0 1 and so 1 > (0). As the notation suggests 1 is the multiplicative inverse of . This is stated in the following. Lemma 7. If is a real number and > (0), then 1 = (1).
Before we prove this we shall record a technical consequence of Lemma 4 of Chapter 2 that we shall be using. Lemma 8. If is a real number, then for any counting number n, there exist rational numbers a and b, b , a such that b a < 1 and there exists a rational n number c such that c < b.
28 Ng Tze Beng
This lemma simply says that we can always choose the rational number b there not equal to the minimum of the complement of in Q if this minimum exists. Proof of Lemma 8. By Lemma 4 of Chapter 2, for any counting number n there exist rational numbers a and c, a , c such that 0 < c a < 1 . Take b = 2c 2n a. This choice of b satisfies b > c and so b and b a = 2c 2a < 1 . n Proof of Lemma 7. Recall that the product
1 = (0) { x Q: x = ab for some rational number a 0 in and some rational number b 0 in 1 }. 1 (1). Since (0) (1) and 0 (1), we only need to show that for any rational number a > 0 in and any rational number b > 0 in 1 , We shall show first that a b (1). By the definition of 1 , b > 0 in 1 implies that there exists rational number y such that (0 < ) b < 1/y. Since a and is a real number, a < y. Therefore, a b < y (1/y) = 1 and so a b (1). Thus 1 (1) and that means 1 (1). Next we need to show that (1) 1 . Now for any rational number x 0, x 1 since (0){0} 1 because 0 belongs to both and 1 and (0) 1. 1.
Then it remains for us to show that for any rational x such that 0 < x < 1, x Choose a rational number y 1 . Then for all rational number z ,
0
y0 1/z ( > 0). -------------------------------------------- (1) Next since Q is Archimedean, there exists a counting number n such that n (1 x) > y0, that is to say, 1 1 y 0 > x. n ---------------------------------------------- (2)
By Lemma 8, there exist rational numbers x0 and z0, where x0 and z0 , such that 0 < z0 x0 < 1 ---------------------------------------------- (3) n and there is a rational number z1 such that (0 < ) z1 < z0 . Then 1/z0 < 1/z1 and x0 1 so 1/z0 1 . Dividing (3) by z0, we have 0 < 1 z 0 < nz 0 . Therefore, x0 1 z 0 > 1 nz 0 . --------------------------------------- (4)
29 Ng Tze Beng
1 . By (1), y m 1 since z . Thus 1 m y 0 . 0 0 z0 nz 0 n x0 1 y > x. Thus x 1 . This is true for Therefore, by (4) and (2) we have z 0 > 1 n 0 all x < 1. Therefore, (1) ` 1 and so (1) [ 1 . Thus we have proved that x0 1 Note that z 0 = x 0 z 0 c 1 = (1). Next we shall verify that our multiplication on the real numbers R has the desired properties at least for the moment for the real numbers (0). Lemma 9. (Properties of Multiplication) Let , and be any real numbers (0). Let a and b be any rational numbers 0. Then 1. 2. 3. 4. 5. 6. (0); if and > (0), then > (0). = . ( ) = ( ). ( + ) = + . (1) = . (a b) = (a) (b). (Commutativity) (Associativity) (Distributivity)
Property 1 is an immediate consequence of the definition as explained just after the Definition 3. Property 2 is Proposition 4 part 1 and Property 3 is an easy consequence of the definition of the multiplication and the associativity of multiplication on the rational numbers. We shall prove the remaining three properties. Proof of Property 5 of Lemma 9. Observe that (1) = (0) {x Q: x = ab for some rational number a 0 in and some rational number b 0 in (1)}. Since (0), (0) . Now for any a 0 in and any rational number b 0 in (1), a b a since b < 1. Thus a b since a and is a real number. This implies that {x Q: x = ab for some rational number a 0 in and some rational number b 0 in (1)} . Hence (1) = (0) {x Q: x = ab for some rational number a 0 in and some rational number b 0 in (1)} . That means (1) . Next we shall show that (1). Let x be a rational number in . If x < 0, then x (0) (1). If x 0, then by Lemma 4 Part 2 of Chapter 2, there exists a counting number n such that x + 1 . Obviously x + 1 > x 0. Thus 0 x/(x + 1 ) < 1 and n n n so x/(x + 1 ) (1). Hence x = (x + 1 )(x/(x + 1 )) (1). Therefore, we can n n n conclude that (1). We can now conclude that = (1).
30 Ng Tze Beng
Proof of Property 4 of Lemma 9. First observe that and (0) implies that + (0) by Lemma 1 Part 4 of Chapter 3. Similarly we conclude that + (0). We shall show that ( + ) + . Take a rational number x in ( + ). If x < 0 then x (0) + . We now assume that x 0. Then there exists rational number a 0 in and rational number b 0 in ( + ) such that x = a b. Since b ( + ), there exist rational numbers c in and d in such that b = c + d. Therefore x = a b = a (c + d) = a c + a d. Since c + d = b 0, either c 0 or d 0. If both c and d are 0, then a c and a d and so x = a c + a d + . If c 0 and d < 0, then c + d < c and so b = c + d since c is in . Thus x = a b . ow = + (0) + by Lemma 2 Part 4 of Chapter 3 since (0). herefore, x + . Likewise, if c < 0 and d 0, then c + d < d and so b = c + d since d is in . Hence as before, we have x = a b + . Once more we get x + . Thus we can conclude that ( + ) + . Next we show that + ( + ). By Lemma 1 Part 4 of Chapter 3, + , since and (0). We have thus + and + . Then ( + ) and ( + ). This is seen as follows. Take any x in . If x < 0, then x ( + ). If x 0, then there exist rational number a 0 in and rational number b in such that x = a b. Since + , b + . Therefore, x = a b ( + ). Hence ( + ). It follows that ( + ). Now take any rational number x in and any rational number y in . If x < 0, then x + y < y and so x + y ( + ) since is a real number. Likewise if y < 0, then x + y < x and so x + y ( + ). If both x and y are 0, then there exist rational numbers a, a' 0 in , b 0 in and b' 0 in such that x = a b and y = a' b' . If a' a, then x + y = a b + a' b' a b + a b' = a(b + b' ) ( + ). Hence x + y ( + ) because ( + ) is a real number. Finally and similarly if a a' , then x + y a' (b + b' ) ( + ) and so we deduce in the same manner that x + y ( + ). Since this is true for any x in and any y in , we conclude that + ( + ). Therefore, + = ( + ). Proof of Property 6 of Lemma 9. If either a or b is 0, then (a b) = (0) and (a)(b) = (0) by Proposition 4 Part 2. In this case the equality is trivial. We now assume both a and b are > 0. Let y be any rational number in (a)(b). If y < 0, then y < a b and so y (a b). We now assume y 0. Then there exist rational numbers c and d such that 0 c < a and 0 d < b and y = c d. Therefore, y = c d < a b and so y (a b). Hence any rational number y in (a)(b) is also in (a b). Thus (a)(b) (a b). Conversely take any rational number x in (a b). If x 0, then x x (a)(b) since (a)(b) > (0). We now assume 0 < x < a b. Then 0 < a < b since a x > 0. Then by the Archimedean property of Q, there exists a > 0. Therefore b a x x x x counting number n such that n (b a ) > a . Hence (b a )> 1 a . Thus n x b > (1 + 1 ) a . n x Therefore, (1 + 1 ) a (b). Also since a 1 < a, a n 1+ n 1+
1 n
(a). Hence
31 Ng Tze Beng
x x = ( a 1 ) ( (1 + 1 ) a ) (a)(b). We can now conclude that (a b) (a)(b). n 1+ n Thus (a b) = (a)(b). This completes the proof.
Extension of Multiplication to All of the Real Numbers

Extending the multiplication to all of the real numbers R is a natural progression for our construction. We shall use the device: convert a pair of real numbers, if needed, to a pair of real numbers (0) by taking the appropriate additive inverse, then multiply the resulting pair according to the definition for real numbers (0), follow this by, if appropriate, taking the additive inverse of the real number so obtained. For real number < (0), the operation of taking the additive inverse behaves like the usual reflection for the rational numbers. This is made precise in the following lemma. This lends support to the economy of visualisation of the real numbers as an infinite "line". Lemma 10. For any real number , < (0) if and only if > (0). Proof. Suppose < (0). Then (0) and there exists a rational number y < 0 such that y . Take any rational number z in (0). Then z < 0 < y. This means that z and 0 belongs to . his implies that (0) and that means (0) < since 0 . Conversely suppose (0) < . hen (0) and there exists a rational number y0 in such that y0 (0). Hence y0 0 and there exists some rational number y1 such that 0 y0 < y1 . Therefore y1 < y0 0. Since y1 , for any rational number x in , x < y1 < 0. Thus (0). Since y1 (0) and y1 , < (0). his completes the proof. Definition 11. Suppose , and are real numbers such that < (0), < (0) and (0). The following define the multiplication of with any one of or in any order. = (( ) ) = ( ( )) = and = ( )( ) = ( )( ) = . Observe that with Definition 11, multiplication of any two real numbers not both (0) is defined. This means that we now have a multiplication on the set of real numbers. A consequence of the definition is commutativity for multiplication. Another deduction from the definition is that for any real number , (0) = (0) = (0). With this definition, we have a multiplication map, R R R, that takes an ordered pair (, ) and assign to it the product . It now remains to define the multiplicative inverse of any real number < (0). Definition 12. For any real number < (0), define the multiplicative inverse by 1 = 1 .
32 Ng Tze Beng
We can easily check that this is really the multiplicative inverse of . To facilitate the checking we use the following Lemma. Lemma 13. For any real number , ( ) = . Proof. = + (0) = + (() + (( ))) = ( + ()) + (( )) by the associativity of addition = (0) + (( )) = (( )) + (0) by the commutativity of addition (Lemma 2(1) Chapter 3) = ( ) by Lemma 2 part 3 of Chapter 3.
Thus for < (0), > (0) and so
> (0) as remarked in the paragraph
following the definition of the multiplicative inverse. It follows then that 1 = 1 < (0) by Lemma 10. Hence by Definition 11, 1 = ( ) 1 = ( ) 1 = ( ) 1 by Lemma 13 by Definition 12.
= (1) by Lemma 7. Thus the definition of the multiplicative inverse of the real number < (0) is the correct one.
Now we collect together all the properties R will have with respect to multiplication and addition and which together with the properties in Lemma 2 of Chapter 3 characterise it as a field. Lemma 14. For any real numbers , and and any rational numbers a and b, the following properties hold. 1. = . 2. ( ) = ( ). 3. 4. ( + ) = + . (1) = . 1 = (1). (Commutativity) (Associativity) (Distributivity)
5. If (0), then
6. If > (0), then > > . 7. (a) (b) = (a b).
33 Ng Tze Beng
Property 1 of Lemma 14 is a consequence of Lemma 9 part 1 and Definition 11. Property 5 has already been shown. We shall prove the remaining properties. Proof of Property 2 of Lemma 14. If any one of , or is (0), then trivially both sides of the equation are the same as (0) and so we have nothing to prove. If , , > (0), then this is just Lemma 9 Part 2. So we have the remaining seven cases to check out. Case 1. , > (0), < (0). ( ) = ( ( )()) since > (0) and < (0) = ( ( ( ))) by Lemma 9 Property 3 = = (( ( ))) by Definition 11 since > (0) and ( ( )) < (0) and Lemma 13 ( ) by Definition 11 since > (0) and < (0) .
Case 2. , > (0), < (0). ( ) = ((( ))) by Definition 11 since > (0) and < (0) = ( (((( ))))) by Definition 11 since (( )) < (0) and > (0) = ( (( ))) by Lemma 13 = ( (( )) ) by Lemma 9 Property 3 since , , > (0) = ( (( ))) by Definition 11since > (0) and (( )) < (0) and Lemma 13 = ( ) by Definition 11 since < (0) and > (0). Case 3. > (0), , < (0). ( ) = ((( ))) by Definition 11 since > (0) and < (0) = ( ((( ))))() by Definition 11 since < (0) and (( )) < (0) = (( ))() by Lemma 13 = (( )()) by Lemma 9 Property 3 since , , > (0) = ( ) by Definition 11 since , < (0). Case 4. < (0), , > (0). ( ) = ( ) by commutativity of multiplication = ( ) by Case 2 above = ( ) by commutativity of multiplication = ( ) by Case 1 above = ( ) by commutativity of multiplication.
34 Ng Tze Beng
Case 5. , < (0), > (0). ( ) = ( ) by commutativity of multiplication = ( ) by commutativity of multiplication = ( ) by Case 3 above = ( ) by commutativity of multiplication = ( ) by commutativity of multiplication. Case 6. , < (0), > (0). ( ) = ( ) by commutativity of multiplication = ( ) by Case 5 above = ( ) by commutativity of multiplication = ( ) by Case 5 above = ( ) by commutativity of multiplication. Case 7. , , < (0). ( ) = (()()) by Definition 11 since , < (0) = (( ()())()) by Definition 11 since < (0) and ()() > (0) = (()(()())) by Property 3 of Lemma 9 since since , , > (0) = (()()) by Definition 11 since < (0) and ()() > (0) = ( ) by Definition 11 since , < (0). This completes the proof of Property 2 of Lemma 14. Proof of Property 4 of Lemma 14. If (0), then (1) = by Property 5 of Lemma 9. For real number < (0), (1) = (( )(1)) by Definition 11 = ( ) by Property 5 of Lemma 9, since > (0) = by Lemma 13. The following is a useful observation of the operation of taking additive inverse with multiplication. It is an easy consequence of Definition 11. It is stated here for reference since we shall be using it implicitly. Lemma 15. For any real numbers and , (()) = ()() = (()) = . Proof of Property 3 of Lemma 14.
35 Ng Tze Beng
If , , (0), then ( + ) = + by Property 4 of Lemma 9. If = (0), trivially, ( + ) = + = (0) by Proposition 4. If = (0), then ( + ) = ((0) + ) = = (0) + = + . Similarly, if = (0), then ( + ) = + . The remainder of the proof is covered in the following six (not necessarily mutually exclusive) cases. Case 1. > (0), (0), + (0), If (0), then the distributivity is already shown in the preceding paragraph. We now assume < (0). ( + ) + ( ) = (( + ) + ( )) by Lemma 9 Property 4 since + , (0) = (( + ( + ( ))) by associativity of addition (Property 2 of Lemma 2 Chapter 3) = ( + (0)) by Property 3 Lemma 2 of Chapter 3 = . Adding (( )) to both sides, we get (( + ) + ( )) + ((( ))) = + ((( ))) = + by Lemma 15. Thus ( + ) + (( )) + ((( )))) = + by the associativity of addition. Hence the left hand side is equal to ( + ) + (0) = ( + ). This proves ( + ) = + . Case 2. > (0), (0), + (0). ( + ) = ( + ) by the commutativity of addition (Lemma 2 Property 1 of Chapter 3). = + by Case 1 above, = + by the commutativity of addition. Case 3. > (0), + < (0). ( + ) = ((( + ))) by Lemma 15 = ((() + ())) by Lemma 4 Part 3 of Chapter 3 = (() +()) by Case 1 if (0) or Case 2 if (0) since > (0), and () + ()> (0) = (()) + ((())) by Lemma 4 Part 3 of Chapter 3
36 Ng Tze Beng
= + by Lemma 15. Case 4. < (0), (0) and + (0). ( + ) = (()( + )) by Lemma 15 = (() + () )) by Case 1 since > (0), (0) and + (0) = (() ) + ((())) by Lemma 4 Part 3 of Chapter 3 = + by Lemma 15. Case 5. < (0), (0) and + (0). ( + ) = ( + ) by the commutativity of addition = + by Case 4 above = + by the commutativity of addition (Lemma 2 Property 1 of Chapter 3). Case 6. < (0), + < (0). ( + ) = (()( + )) by Lemma15 = (() + ()) by Case 3 since > (0) and + < (0) = (()) + ((())) by Lemma 4 Part 3 of Chapter 3 = + by Lemma 15. This completes the proof of Property 3. Before we proceed to prove the remaining properties, we state the following equivalent result of Property 4 Lemma 2 of Chapter 3 for strict inequality. Lemma 16. For any real numbers , and , < + < + . Proof. By Property 2 Lemma 2 of Chapter 3, < + + . We need only show that there exists a rational number d in + not in + . Now since < , there exists a rational number a in , but a . Therefore, for all x in , x < a. By Lemma 4 Part 2 of Chapter 2, there exists a counting number n such that a + 1/n . Also by Lemma 4 Part 1 of Chapter 2, there exists rational numbers b in and c such that c b < 1/n. Therefore, for all rational number y in , y < c < b + 1/n. Thus for all rational number x in and all rational number y in , x + y < a + b +
37 Ng Tze Beng
1/n. It follows that d = a + b + 1/n + . Now d = (a +1/n) + b + since a +1/n and b . We can thus conclude that + < + . Proof of Property 6 of Lemma 14. By Lemma 16, > implies that + () > + () = (0). Since > (0), by Property 1 of Lemma 9, ( + ()) > (0). Therefore, by Property 3 of Lemma 14, we have + () = ( + ()) > (0). It then follows by Lemma 16 that ( + ()) + (( ())) > (0) + (( ())). Now (0) + (( ())) = (( ())) by Lemma 2 Property 1 and 3 of Chapter 3 = by Lemma 15. Also ( + ()) + (( ())) = + ( () + (( ()))) by Property 2 Lemma 2 of Chapter 3 = + (0) = by Lemma 2 Property 3 of Chapter 3. Hence > . Proof of Property 7 of Lemma 14. If a = 0 or b = 0, then trivially (a) (b) = (a b) = (0). If a, b > 0, then (a) (b) = (a b) by Property 6 of Lemma 9 . If a > 0 and b < 0, then a b < 0. Therefore, (a) (b) = ((a) ((b))) = ((a) (b)) by Lemma 4 Part 2 of Chapter 3 = (a (b)) by Lemma 9 Property 6 since a, b > 0 = ((a(b))) by Lemma 4 Part 2 of Chapter 3 = (a b). If a < 0 and b > 0, then (a b) = (b a) = (b) (a) by the above argument. Thus (a b) = (b) (a) = (a) (b) . If a, b < 0, (a) (b) = ((a))( (b))
38 Ng Tze Beng
= ( a)(b) by Lemma 4 Part 2 of Chapter 3. = (( a)(b)) by Lemma 9 Part 6 since a, b > 0 = (a b). This concludes the proof of Lemma 14 Part 7. We next extract some properties of the real numbers with respect to multiplication. We summarise this in the following theorem. Theorem 17. R {(0)} is an abelian group with respect to multiplication and its identity element is (1). That is multiplication on R is commutative and associative, has an identity element (1) and a unary operation that assigns to each real number (0), its multiplicative inverse 1 . This is a consequence of Lemma 14 (1), (2) and (4). Lemma 14, Theorem 5 of Chapter 3 and Theorem 10 of Chapter 2 actually imply the following: Theorem 18. R is a field with (0) as the identity element for addition and (1), the identity element for multiplication. R has a total ordering with respect to ' '. R is a complete totally ordered field. Lemma 11 of Chapter 2, Lemma 2 Part 5 of Chapter 3 and Lemma 14 Part 7 says that the embedding of the rational numbers into R, : Q R is a monomorphism of the field of rational numbers into R preserving the ordering on Q. We have now constructed a field R with the rational numbers embedded in R as a subfield. To think of the real numbers R as an extension of the rational numbers Q, we identify each rational number a in Q as its image (a) in R. In this way the rational numbers is thought of as the subfield (Q) of R. When we speak of the rational numbers, we think of the usual rational numbers, but when we want to prove any property of the real numbers we use the properties of a complete totally ordered field and think now of the rational numbers as the embedded subfield (Q). It is hard to work with the model of the real numbers that we have constructed; most results about real numbers are results about complete totally ordered field and are proved using the properties of a complete totally ordered field. This is because there is essentially one such complete totally ordered field and we shall discuss this in the next chapter. We conclude this chapter showing that the remark stated just before Lemma 10 is explained as the fact that the additive inverse of any real numbers is precisely the multiplication of this real number by the (rational) number (1), the equivalent of 1 in R. Lemma 19. For any real number , = (1) . Proof. By Lemma 2 Part 5 of Chapter 3, (0) = (1) + (1). Now (0) = (0) = ((1) + (1))
39 Ng Tze Beng
= (1) + (1) by distributivity = + (1). Therefore, adding the additive inverse to both sides of the equation above yields + (0) = + ( + (1) ) = (() + ) + (1) by the associativity of addition = (0) + (1) = (1) + (0) by the commutativity of addition. Thus by Lemma 2 Part 3 of Chapter 3, = (1) . This completes the proof.
40 Ng Tze Beng
CHAPTER FIVE
Order! Order! There is only one order

The ordering on the real numbers R is given by subset inclusion. Let us now examine how we can think of this ordering as given essentially by a positive cone in R as described in Chapter 1. Let P = { R: > (0)}. For any real number (0), > (0) or < (0). By Lemma 10 Chapter 4, < (0) if and only if < (0). Thus the set P = { R: > (0)} = { R: < (0)}. Hence R = P P {(0)}. In particular we have, Theorem 1. 1. (0) P. 2. , P + P. 3. , P P. 4. (0) P or P. Proof. Part 1 is obvious. For Part 2, , P implies that , > (0). It follows by Lemma 16 of Chapter 4 that + > (0) + = > (0). ence + P. Part 3 is just a consequence of Lemma 9 Part 1 of Chapter 3. Part 4 follows from Lemma 10 of Chapter 4 and the fact that the ordering on R is a total ordering. Thus P is a positive cone for the real numbers. The ordering with respect to the positive cone is defined by ">" if and only if P . Now P if and only if > (0) if and only if > + (0) = . So the ordering defined by the positive cone is the same as that defined by set inclusion. So if we define a real number to be positive if P, then the positive cone is none other than the positive real numbers and is positive if > (0) . So we may if we wish identify (0) with 0 but we must be careful to distinguish them. Notice then that a positive cone for the embedded rational number (Q) is P (Q), which is the same as the set {(x) : x a rational number and x > 0} and is the image of the positive cone of the rational numbers Q. The uniqueness for the positive cone for the rational numbers Q has been established in Chapter 1. It is none other than the set of positive rational numbers. However, the uniqueness of the positive cone for R and hence the ordering on R is a little harder to show.
Chapter 5 Order! Order! There is only one order.
We now formally set the definition of a positive number and a negative real number. Definition 2. A real number is said to be positive if > (0). is said to be negative if < (0). We shall need the following technical lemma regarding the existence of the n-th root of a positive real number. We require only the existence of the square root of a positive real number to show the uniqueness of the positive cone. Nonetheless, the same proof applies for n-th root. Lemma 3. Let n be any counting number. If is a positive real number, that is > (0), then there exists a positive real number such that n = . Specifically for n = 2, there exists a real number such that 2 = . We shall prove Lemma 3 only for n = 2. First, we make the following definition. Definition 4. For any > (0), define = {x Q: x 2 < k for all k } (0). Obviously 0 . he set above is our candidate for the square root of . Note the construction of the square root has a practical and constructible nature in the set theoretic sense. In the usual way that we talk about solution to x2 = p, assumption is made about the existence of the square root of p without question. Symbol like p is employed to mean the square root of p. The meaning of p is not clear, particularly so when p is a prime number. The meaning and existence of p must be established less we shall be talking about an entity that does not exist. In this context we shall use the embedded rational numbers (Q). Use (p) instead of the prime p. With = (p) and n = 2, the set define above will be the square root of (p). Undoubtedly (Q). Of course, we need to show that is a real number. Lemma 5. number. For any real number > (0), the set defined in Definition 4 is a real
Proof. 1. Since (0) , . 2. Since is a (left) cut, there exists a positive rational number q > 1 not in . Then for any rational number x in , x 2 < q 2 and so x < q. Therefore, q and so Q. 3. Let l and x < l. We shall show that x . If x 0, then x . It is sufficient to show that if x, l > 0 and x < l, then x . Now 0 < x < l implies that x 2 < l 2 and l implies that l 2 < k for all k not in . It follows that x 2 < k for all k not in and so x by Definition 4. 4. Next we shall show that has no maximal number. Suppose on the contrary that has a maximal number M. Take a rational number q > 0 in . This number exists by Lemma 2 of Chapter 4. Since Q is Archimedean, there is a counting number n 2 such that n q > 1, that is 0 < 1 < q. In particular, 12 < q. Therefore, 1 < q < k n n n
42 Ng Tze Beng
for any rational number k not in . Hence 1 and 1 > 0. Thus M > 0. Then M 2 n n . This is because, if on the contrary M 2 , then M would imply M 2 < M 2, which is absurd. Then by Lemma 4 part 2 of Chapter 2, there exists a counting 1 1 number m such that M 2 + m . Let x0 = M 2 + m . Obviously, x0 > M 2 . By the Archimedean property of the rational numbers, there exists a counting number n1 such 1 1 that n1(x0 M 2 ) > M 2 . Thus (x0 M 2) > n 1 M 2 . Therefore, x0 > (1 + n 1 )M 2 > M 2. 1 Repeating this argument with (1 + n 1 )M 2 in place of M 2 we obtain another counting number n2 such that 1 1 1 x0> (1 + n 2 )(1 + n 1 )M 2 > (1 + n 1 )M 2 > M 2. 1 1 Now let k = max {n1 , n2}. Then M 2 < (1 + 1 ) 2 M 2 < (1 + n 2 )(1 + n 1 )M 2 < x0. This k means that (1 + 1 ) 2 M 2 < y for all y not in since x0 and is a (left) cut. Hence k by the definition of , (1 + 1 )M . (The above proceeding works equally well k with the exponent 2 replaced by any other counting number N , taking k = max(n1 , n2 , ..., nN ) , where the counting numbers nj are obtained by repeating the above argument N times and we get MN < (1 + 1 ) N M N < x0.) Now M < (1 + 1 )M but M is k k the maximal number in and so M (1 + 1 )M and we have a contradiction. Thus the k maximal number M does not exist. By Lemma 3 of Chapter 2, is a real number. This completes the proof of Lemma 5. Lemma 6. Let be defined as in Definition 4 for any real number > (0). Then 2 = . Proof. Recall 2 = (0) {x Q: x = a b for some rational numbers a, b 0 in }.
First, we shall show that 2 . Obviously, (0) . Since 0 , 0 2. Observe that 0 because > (0). It remains to show that for any rational number a, b > 0 in , a b is in . Since a, b , for all rational number k , a 2 , b 2 < k. Therefore, for all rational number k , (a b)2 = a 2 b 2 < k 2. This means that a b < k for all rational number k . Thus a b . We can then conclude that 2 . Next we shall show that 2. As before it is enough to show that for all rational number y > 0 in , y 2 since (0) {0} , 2. If on the contrary y 2, then y > a 2 for all rational number a > 0 in , otherwise y a 2 for some a > 0 in would imply that y 2. Therefore, we have for all a in , a 2 < y < k for all k --------------------------- (1)
Now by Lemma 4 of Chapter 2, for any counting number n, there exist rational numbers a , b such that b a < 1/n. Since > (0), b implies that b > 0 and for some rational number q not in , b 2 q. Hence by (1) we get a2 < y < q b2 < (a + 1 )2 ------------------------------- (2) n
43 Ng Tze Beng
Now we are going to prepare y for a contradiction. Since y , there exists a rational number x0 in such that y < x0. Choose a fixed rational number k0 . Then k0 > 0, since > (0). Since x0 y > 0, by the Archimedean property of Q, there exists a counting number n0 such that n0 and hence x0 y > 1 + 2 k0 2 ------------------------------- (*) -------------------------------- (3)
x0 y 1 n 0 (1 + 2k 0 ) < 2 .
With this counting number n0 we obtain rational numbers a in and b not in satisfying (2). That is, 1 a2 < y < b2 < (a + n 0 )2 -------------------------- (4) 1 1 1 1 1 Now, (a + n 0 )2 a2 = n 0 ( n 0 + 2a) < n 0 (1 + 2k 0 ) since n 0 < 1 and a < k0 because a x0 y 1 1 and k0 . Therefore, (a + n 0 )2 < a2 + n 0 (1 + 2k 0 ) < a 2 + by (3). Hence 2 x0 y x0 + y x0 + x0 1 by (4), (a + n 0 )2 < y + = < = x 0 since y < x0. Now since x0 2 2 2 1 , for any rational number k , x0 < k. Therefore, for all k , (a + n 0 )2 < k. This 1 1 1 means (a + n 0 ) . But (a + n 0 ) because (a + n 0 ) > b, b and is a (left) cut. Thus we have a contradiction and we conclude that y 2 . This is true for any rational number y in and so 2 . Thus 2 = . This completes the proof of Lemma 6. For completeness we remark that a candidate for the N-th root of a real number > (0) would be the following real number N = (0) {x Q: xN < k for all rational number k }. The verification that N is a real number is similar to that for N = 2 as indicated in the proof therein. The proof that (N )N = is the same as that for N = 2 except that in place of (*) we use the existence of a counting number n0 such that n0 x0 y > 1 + NC1 k0 + NC2 k02 + + NCN1 k0N1 2
for some fixed rational number k0 . The uniqueness of the ordering on the real numbers is summarised as follows. Theorem 7. The real number field R has exactly one total ordering. That is to say if there is another positive cone P' for R, then P' = P. Proof. Suppose P' is another positive cone satisfying R = P' P' {(0)} and that for any real numbers , P' , + , P'. Take any real number (0),
44 Ng Tze Beng
either P' or P' . If P' , then 2 P'. If P' , then P' and so 2 = () 2 P'. Therefore, for any (0), 2 P'. For any P, > (0) and so by Lemma 6, there exists a real number > (0) such that = 2 . Therefore, P' . Hence P P' . Thus the complement R P' of P' satisfies R P' R P. That is P' {(0)} P {(0)}. Since (0) P' , P' P. It follows then that ( P') ( P) . Because for any subset of real numbers M, ( M) = M by Lemma 13 of Chapter 4, we conclude that P' = ( P') ( P) = P. Thus P' = P and so the total ordering on R is unique. For any complete totally ordered field F, there is a subfield isomorphic to the rational numbers. This is seen as follows. Let 1 denote the identity element with respect to multiplication. Then Q is embedded as the subfield { m1 : m an integer and n a counting number}, n1 where we write 1 for the multiplicative inverse of n1. Let us see if this gives a new n1 embedding of the rational number Q into our real numbers R. The identity element for multiplication is (1). The embedding of the rational number m is given by n m (1) (m) 1 ) = ( m ) which is the same as the original embedding in = (n) = (m) ( n n n (1) 1 Chapter 2. In the above we have use the fact that (n) = ( 1 ). This is seen as follows. n 1 ) = (n 1 ) = (1) by Lemma 14 of Chapter 4. Therefore, Since (n) ( n n
1 1 1 1 1 1 1 1 (n) = (n) (1) = (n) ( (n) ( n )) = ( (n) (n)) ( n ) = (1) ( n ) = ( n ).

The Archimedean property of a complete totally ordered field is a consequence of its completeness. The following is proved similarly as Proposition 2 of Chapter 1. Theorem 8. Any complete totally ordered field is Archimedean. As a consequence of the Archimedean property of a complete totally ordered field F, the embedded rational numbers is dense in it. That is for any , in F and < , there exists a rational number m with m an integer and n a counting number such that < m1 < . The n n1 proof is exactly the same as the proof of Corollary 4 of Chapter 1. We shall be using this fact again and again. We have shown that the real numbers R is a complete totally ordered field. Does there exist a complete totally ordered field very different from R? The answer is no, at least in the sense that though the objects or constructs may be different nevertheless we can change from one object or construct to another for all practical purposes. This is what the next theorem says.
45 Ng Tze Beng
Theorem 9. For each complete totally ordered field F, there is an order preserving isomorphism of F into R and this isomorphism is unique. That is to say there is ONLY one such isomorphism. We shall prove this theorem in stages. First, we shall define the isomorphism as stated in Theorem 9. Lemma 10. For any in a complete totally ordered field F, the subset M = {x Q: x1 < } is a real number in R. (Here we write for x = m , x1 = m1 .) n n1 Proof. 1. M is not empty. This is seen as follows. For any in F, either > 0 the identity element for addition, or < 0 or = 0. If > 0, then 0 = (0)1< and so 0 M and M . If < 0, then by the Archimedean property of F, there exists a counting number n such that n1 >() and so n1= (n)1< and so n M and once more M . If = 0, then any negative rational number is in M since for any negative rational number q, q1 < 0. Obviously for this case M . 2. M Q. If > 0, then by the Archimedean property of F there exists a counting number n such that n 1 > . Thus we have a rational number n not in M. If 0, then for any rational number q > 0, < q1, since q1 > 0. That means q M and so M Q. 3. Take any y in M. Then y1 < . Therefore, for any rational number x < y, we have then x1 < y1 < and so by definition of M, x M. 4. M has no maximal element. Suppose M has a maximal element J. Then for all x in M, x J and so x1 J1 < . Again by a consequence of the Archimedean property of F (density of the embedded rational field), there exists a rational number q such that J1 < q1 < . (Refer to the proof of Corollary 4 of Chapter 1 for verification.) Thus J < q and q M. This contradicts that q J . Therefore, M has no maximal number. We now conclude by Lemma 3 of Chapter 2 that M is a real number. We can now proceed to define the isomorphism. Definition 11. For any complete totally ordered field F, define a mapping : F R by () = {x Q: x1 < }, where 1 denotes the identity element for multiplication in F, for in F. By Lemma 10 this mapping is well defined. Lemma 12. The mapping : F R is an order preserving bijection. Proof. is injective. This is seen as follows. Let and be in F such that . Then either < or > . Without loss of generality we may assume < . By Definition 11, () (). Now by the density of the embedded rational field (see Corollary 4 of Chapter 1) there exists a rational number q such that < q1 < . Then
46 Ng Tze Beng
() < () since q () but q (). This implies that the mapping is order preserving. Thus () () for and so is injective. Take any real number in R. Then consider the set = {x1: x } F. This set is bounded above. This is because there exists a rational number q such that for all x in , x < q since is a cut. Therefore, for all x in , x1 < q1 and so is bounded above by q1 in F. Since F is complete, has a least upper bound or supremum say in F. Then () = . This is seen as follows. Obviously for any rational number y in () , y1 < and so y1 . This is deduced below. If on the contrary y1 then y > z for all z in and so y1 > z1 and y1 would be an upper bound of . Thus y1 contradicting y1 < . Hence y1 and so y . This means () . Obviously () since the above argument can be reversed as follows. For any rational number x in , x1 . Thus x1 since is the least upper bound of . We claim that x1 . This is because if x1 = , then for all y in , y1 x1. Thus for all y in , y x. Therefore, x would be a maximal number for contradicting that has no maximal number. Thus for all x in , x1 < and so x (). This implies that (). Thus () = . This shows that is surjective and so is bijective. Lemma 13. The mapping : F R is a field homomorphism. That is for any and in F, 1. ( + ) = () + () and 2. ( ) = () (). Proof. 1. First, we shall show that () + () ( + ), that is () + () ( + ) . For any x in (), x1 < and for any y in (), y1 < . Therefore, for any x + y in () + () with x () and y (), (x + y)1 = x1 + y1 < + . Hence x + y ( + ). Therefore, () + () ( + ). Now take any z in ( + ). Then z1 < + and so z1 < . Then by the density of the embedded rational field in F, there exists a rational number q such that z1 < q1 < . Then we have (z q)1 = z1 q1 < and so z q (). Obviously q (). Thus z = (z q) + q () + (). This is true for any z in ( + ) and so ( + ) () + (). Therefore, () + () = ( + ). 2. First, we note that (0) = (0) the identity element for addition in R. As a consequence of part (1) above, for any in F, () = (). (This can be deduced in a similar manner as in the proof of the corresponding fact for the embedding of the rational numbers into R.) Note also that (1) = (1). If any one of and is 0, then we have nothing to prove since ( ) = (0) = (0) and since ()() is either ()(0) = ()(0) or (0)() =(0)() and is always the same as (0). It is sufficient to show Part 2 for and not equal to 0 the identity element for addition in F. We shall divide this into two cases, first for , > 0, then for either > 0 and < 0 or > 0 and < 0 or , < 0. Now assume , > 0. Note that for any > 0 in F, () > (0) and so by Lemma 2 of Chapter 4, {0} (0) (). Now (), () , ( ) > (0). Then we have ()() > (0). Hence {0} (0) ()(), ( ). So for any x 0 in
47 Ng Tze Beng
()(), x ( ). Therefore, we only need to show that for any x > 0 in () and any y > 0 in (), x y ( ). Now x > 0 in () implies that x1 < and likewise y > 0 in () implies that y1 < . Therefore, (x y)1 = x1 y1 < and so x y ( ). We can now conclude that ()() ( ). Next we shall show that ( ) ()(). Again we need only show that for any z > 0 in ( ), z ()(). (The proof of this part resembles that of the proof for Lemma 9 Part 6 of Chapter 4.) Now since z > 0 and is in ( ), 0 < z1 < . Therefore, multiplying the inequality by the multiplicative inverse of , we get 0 < z1 1 < . By the density of the embedded rational field in F, there exists a rational number a such that 0 < z1 1 < a1 < .
z z Then a () and 0 < z1 < a1 < . Thus 0 < z1 = a 1 < and so a () . a1 z Therefore, z = a a ()(). This is true for any z in ( ) and so ( ) ()() and the equality ( ) = ()() follows.
Now for > 0 and < 0, () > (0) and () < (0). Therefore, by the preceding remark, ()() = (()(())= ((()(()) = (()) by the first case since , () > 0. Thus ()() = ((())) = ( ). Similarly if < 0 and > 0, by interchanging with , we can show that ()() = ()() = ( ). Finally when , < 0, then ()() = (())(()) = ()() = (()()) = ( ) by the first case. This completes the proof of Part 2 of Lemma 13. Lemma 12 and Lemma 13 imply that there is only one complete totally ordered field upto isomorphism. The fields F and R may be constructs of different nature but they are all isomorphic in the sense that the mapping :F R is a bijective homomorphism, an isomorphism. More is true, that is there is one and only one isomorphism of any complete totally ordered field F into R. This is a consequence of the following theorem which is of some interest on its own. Theorem 14. Any automorphism : R R (i.e. isomorphism of R into itself) is the identity homomorphism. Proof. We observe that any automorphism would be order preserving. (x2) = (x)(x) = ((x))2 > 0 if x 0. Here we write 0 for (0). Now for any in the positive cone P, = 2 for some and so () = (2) > 0 and so () P. That means maps positive cone into positive cone. Hence if , R and > , then P and so () () = ( ) P and () >(). We shall write 1 for (1) in R. Then n1 = n(1) = (n) for any integer n. For any counting number n, ( 1 )= 1 = 1 and so for any rational number m , ( m ) = (m) ( 1 ) = m1 . So we n n n n (n) n1 n1 m1 : m is an integer and n a counting number}. can write (Q) = { n1 We shall show next is the identity homomorphism on the embedded rational number fields (Q). (1) = 1 since is a field homomorphism. Then it follows that (1) = (1) = 1. Therefore, for any integer n, (n) = n1. This means that is the identity homomorphism on the embedded integers. Now for any rational numbers, m , with m an integer and n a counting number, m1 = n n1
48 Ng Tze Beng
1 = m1 1 = m1 , assuming that 1 = 1 . We shall deduce this n1 n1 n1 n1 n1 n1 = (n1) 1 = n1 1 . Therefore, assunption as follows. 1 = (1) = n1 n1 n1 1 . Thus is the identity multiplying by 1 on both sides we get 1 = n1 n1 n1 homomorphism on the embedded rational field. Now let be a real number in R and (Q). Suppose on the contrary that () , then either () > or () < . Suppose () > . Then by the Archimedean property of R, there exists a counting number n such that (m1) n(() ) > 1. Thus 1 < () . Now by Corollary 12 of Chapter 2, there exist rational numbers n1 a and b such that (a) < < (b) and (b a) = (b) (a) < ( 1 ). In our present n notation, we have a1 < < b1 and b1 < a1 + 1 . Since is order preserving and is n1 the identity homomorphism on the embedded rational subfield, we have () < (b1) = b1 < a1 + 1 < + 1 . hus we get () < 1 contradicting () n1 n1 n1 > 1 . Similarly if () < , by the Archimedean property of R, there exists a n1 counting number m such that m( ()) > 1 and so 1 < () . With this m1 counting number m, again by Corollary 12 of Chapter 2, there exist rational numbers a and b such that a1 < < b1 and b1 < a1 + 1 . Then we have < b1 < a1 + 1 . m1 m1 Applying the automorphism to the inequality a1 < , we get a1=(a1)< (). Thus < a1 + 1 < () + 1 and so () < 1 contradicting () > 1 . m1 m1 m1 m1 Therefore, () = and we conclude that is the identity homomorphism on R. This completes the proof . Theorem 15. There is exactly one field isomorphism from any complete totally ordered field F into R. Proof. Suppose there are two isomorphisms , : F R. Let 1: R F be the inverse function of . Then 1 is also an isomorphism. This is seen as follows. For any , in R, (1() + 1()) = (1()) + ( 1())) = + = ( 1( + )). Since is injective we have then 1() + 1() = 1( + ). Similarly, (1() 1()) = (1()) ( 1()) = = ( 1( )). Again since is injective we have then that 1() 1() = 1( ). Therefore, 1: R F is a field isomorphism since 1 is also bijective. Then the composite isomorphism 1 : R R is an automorphism. Therefore, by Theorem 14 1 = 1R : R R is the identity homomorphism. Then = 1R = ( 1) = ( 1 ) = 1F = . This completes the proof of Theorem 15.
49 Ng Tze Beng
Proof of Theorem 9. and Theorem 15.
This is now an easy consequence of Lemma 12, Lemma 13
In the next chapter we shall trace the genealogy of the real numbers starting from the natural number system. We shall see how special and unique is the real number system.
50 Ng Tze Beng
CHAPTER SIX Genealogy

Like the drifting seed in an ocean Asking by whom has cast its existence Whence we came, whimpering Who are we? What are we? Complicated, complex and perplexing questions That has no answer in what or who we are Only what we do Even then only Silence shall answer
We shall now trace the genealogy of the real numbers starting from the indivisible atomic individual the empty set. We shall observe the underlying theme, the first construct, the set of natural numbers (which now for this discussion includes the number 0) is unique in the sense that there is only one such natural number system up to isomorphism, a consequence of the Dedekind's Recursion Theorem, and that each embedding of the natural number system and subsequent constructs into the next is unique. The following chart depicts the lineage of the real numbers.
N N N
Q Z (Z - {0})
It is clear that the construct at each stage involves a new kind of object with the embedded starting object taking on a new and more complicated form. Let us start with VON NEUMANN's definition of the natural number system. With the axiom of existence, we have the existence of an empty set. The number 0 is defined to be the empty set , 1 is defined to be the set {}, 2 is defined to be {, {}} and in general n + 1 is defined to be n{n}. It has two things going for it. Firstly the number n has exactly n elements. Secondly the ordering "<" is given by the relation "" ( "is an element of" ). For instance 3 is defined
Chapter 6 Genealogy
by 2 {2} = {, {}}{{, {}}} = {, {}, {, {}}}. Then , {}, {, {}} {, {},{, {}}}. That is 0, 1,2 < 3. And for n > 0, 0,1,,n1 n and so 0,1,, n 1 < n. Addition and multiplication can be defined by using the successor function, S(n) = n {n} = n + 1 as follows. Addition can be defined recursively starting from m + 0 = m, by m + S(n) = S(m+n) and multiplication by m 0 = 0 and m(n+1) = mn + m. Then we can easily check that all the well-known rules of addition and multiplication are satisfied. The next construct is the set of integers. To do this we shall need the notion of a Cartesian product and an equivalence relation as explained below. Definition 6.1. Let A and B be two sets that may or may not be the same. An ordered pair (a, b) where a is an element in A and b an element in B is defined to be the set {{a}, {a, b}}. That the element a comes first before b is determined by the set inclusion {a}{a, b}. The Cartesian product of A and B, denoted by A B is then the set consisting of all the ordered pairs, that is A B = {(a, b): a A, b B}. In our usual sense of direction, for two natural numbers m and n, if m > n, then we would start from n and proceed to get to m by a sequence of application of the successor function S. The number of times that we apply S would give us a natural number. We can think of this as applying the successor function S to the number 0 in m + 0 while keeping m fixed. If we start from m and get to n then this would indicate an opposite direction. The direction is determined by the ordering m > n. So going from m to n indicates a negative direction and hence gives rise to the "negative" of going from n to m. So we would need an ordered pair (n, m) in order to take into account this sense of direction and once we have this we do not even need the ordering of m and n. So we shall start with the Cartesian product of the natural number N N. There is one complication. If m > n, then m + a > n + a. Thus both pairs (n, m) and (n + a, m + a) would represent the same number. We would want to use only one representative from the set {(n + a, m + a): a N}. So we would need a way to distinguish different sets as representing different numbers. So we would want to carve up the Cartesian product N N into subsets, each of which represents a number. Elements in the set {(n+a, m+a): a N} are related by the same difference. More precisely (a, b) and (c, d) are related if and only if b a = d c if we have defined subtraction but since we have not yet define subtraction, (a, b) and (c, d) are related if and only if a + d = b + c. We shall need then the next concept in set theory, namely that of a relation and in particular an equivalence relation. Definition 6.2. Let A and B be two sets. A relation from A to B is a subset R of the Cartesian product A B. For an element a in A and an element b in B we say a is related to b (and we write aRb) if and only if (a, b) is in R. If A = B, then we say the relation R is a relation on the set A Definition 6.3. Suppose R is a relation on a set A . 1. R is reflexive if for all a in A, aRa.
52 Ng Tze Beng
Chapter 6 Genealogy
2. R is symmetric if for any a and b in A, aRb if and only if bRa. 3. R is transitive if for any a, b, c in A, aRb and bRc implies aRc. We say R is an equivalence relation if R is reflexive, symmetric and transitive. Take an element a in A, the equivalence class of [a] consists of all elements in A that are related to a. That is [a] = {b A: bRa}. We can easily show that [a] = [b] if and only if aRb. This would then carve up the set A into disjoint equivalence classes because [a][b] = if a is not related to b. A then becomes the disjoint union of these equivalence classes. We denote the set of all equivalence classes by A/R. We now take A to be the Cartesian product N N. Define a relation ~ on N N by (a, b) ~ (c, d) if and only if a + d = b + c. Thus our relation is the subset of (N N) (N N) given by {((a,b), (c,d)) (N N) (N N): a + d = b + c }. We then have a new way of looking at the natural numbers. We can easily show that ~ is an equivalence relation. Denote the equivalence class of (a, b) by [a, b]. Then 0 would correspond to [0,0] = [1,1], 1 would correspond to [0,1] = [1,2] and in general any natural number n would correspond to [0, n]. The set of equivalence classes N N /~ is then defined to be the integers Z. Addition and multiplication are defined by [a, b] + [c, d] = [a + c, b + d] and [a, b] [c, d] = [ad + bc, ac + b d]. We can easily check that both addition and multiplication are meaningful, that is if we take different representatives from each class of [a, b] and [c, d] and follow the definition we should end up in the same equivalence class. Observe that both addition and multiplication are commutative and associative. The identity element for addition is [0,0] and the identity element for multiplication is [0,1]. Then Z is an abelian group with respect to addition. The additive inverse [a, b] for any [a, b] in Z is [b, a] because [a, b] + [b, a] = [a + b, b + a] = [0, 0]. Z has no zero divisor because for any x and y in Z, x y = [0,0] if and only if x = [0,0] or y = [0,0] the zero number of Z. We can now embed the natural numbers N into Z, by g: N Z defined by g(n) = [0,n] for each n in N. The order relation is then defined by the positive cone P = {[0, n] : n N and n 0}. For any x and y in Z, x < y if and only if y x P. Indeed, the embedding g: N Z respects addition, multiplication and the orderings on both N and Z. If we now identify [0, n] with n, then since the additive inverse [a, b] = [b, a], we have then the negative integer, n identified with [n, 0]. Then Z is none other than the embedded natural numbers and their negatives. The positive cone can then be written as {n: n g(N) and n 0}. The next construct is the system of rational numbers. Now the symbols a and c , we d b can take them as ratios for the time being, represent the same rational number if a d = b c. If a is not equal to c or if b is not equal to d, then these are really distinctively looking ratios even when a d = b c but we consider them as the same. This brings us to the use of an equivalence relation again. Since there are pairs involved and the denominator rightly should not be zero, we consider the Cartesian product Z (Z {0}), where we write 0 for [0,0]. For (a, b) and (c, d) in Z (Z {0}) we define (a, b) ~ (c, d) if and only if a d = b c. This relation is determined by the subset {((a,b), (c,d)) (Z (Z - {0})) (Z ( Z - {0})): a d = b c } of (Z (Z - {0})) (Z ( Z - {0})). We can easily show that this relation is an equivalence relation and we denote the equivalence class of (a, b) by a . Define the b rational numbers to be the set of all equivalence classes, that is Q = Z (Z {0}) /~ .
53 Ng Tze Beng
Chapter 6 Genealogy
Addition is defined by a + c = ad + bc and multiplication by a c = ac . b d bd b d bd 0 and the identity element for multiplication is Then the identity element for addition is 1 1 . Addition and multiplication are easily shown to be commutative, associative and obey 1 the distributive law. Obviously the additive inverse of a is a and the multiplicative b b a ! 0 is given by b . Then Q with these two operations of addition and inverse of a 1 b multiplication is a field. The positive cone for Q, as explained in Chapter 1, is P = { a : a, b b > 0} and gives rise to a total ordering on Q given by, for any x, y in Q, x > y if and only if x y P. Thus Q is a totally ordered field. The embedding h: Z Q given by h(x) = x 1 embeds the integers Z in Q. This embedding is unique, respects the addition, multiplication and the orderings on Z and Q. Finally the last stage is the construction of the real numbers as described in Chapter 2, involving Dedekind cuts. Thus the set of real numbers sits on top of a heap of other constructs. Its nature is as yet unexplained as the empty set itself. We might say the real number system has no nature but that the logical dynamic process of its creation comes close to being described as its nature. The end products say much less than the process in arriving at these products. In Chapter 5, we have seen that the real number system contains the n-th root of any positive rational number for any counting number n. This was a constructive demonstration. We can divide the real numbers into two classes of numbers, the algebraic numbers and the transcendental numbers. The algebraic numbers are those real numbers that satisfy a polynomial with coefficients in the rational numbers whereas a transcendental number does not. Now a Cantorian argument shows that the algebraic numbers can be at most countably infinite, a consequence of the result that the union of countably infinite family of countable infinite sets is countable. The union of two countable sets is countable again irrespective of whether the sets are infinite or not. An easy argument below will then prove that the set of transcendental numbers is uncountably infinite if we can show that the set of real numbers is uncountably infinite. If the set of transcendental numbers is countable, then the set of real numbers being the union of the algebraic numbers and the transcendental numbers, would then be the union of two countable sets and is therefore countable contradicting that the set of real numbers is uncountable. Hence the set of transcendental numbers is uncountably infinite. We shall next show that the set of real numbers is uncountably infinite by employing a very simple argument of Cantor. First, we shall need a common representation of the non-zero real numbers by non-terminating decimals. 6.4. Decimal Expressions. We shall redefine each real number > 0 as the supremum of a sequence of numbers. Take any > 0 in R. Choose the greatest integer a0 < . We can verify its existence as in Corollary 6 of Chapter 1. Then we have a0 + 1 . Consider next the numbers a0 + 0 , a0 + 10 1 , a + 2 , a + 3 , a + 4 , a + 5 , a + 6 , a + 7 , a + 8 , a + 9 . Then take the 10 0 10 0 10 0 10 0 10 0 10 0 10 0 10 0 10
54 Ng Tze Beng
Chapter 6 Genealogy
a largest of these which is < . That is we choose a1 in {0,1,2,,9} such that a0 + 1 < a0 10 a a + 1 + 1 . Let d1 = a0 + 1 . Then d1 < d1 + 1 . Now choose integer a2 , 0 a2 9, 10 10 10 10 a2 a2 1 . Let now d = d + a 2 = a + a 1 + a 2 . We such that d1 + 2 < d1 + 2 + 2 2 1 0 10 10 10 10 2 10 2 10 1 . Continuing like this we shall obtain a sequence a , a , a , , of have d2 < d2 + 2 0 1 2 10 an a1 a2 an integers, 0 ai 9, such that if dn = dn1 + n = a0 + + + + n then dn < 10 10 10 10 2 1 . The symbol a a a a is called the non-terminating decimal expansion of dn + n 0 1 2 3 10 . So the set {d1 , d2 , d3 , d4 } is bounded above by . Then the least upper bound or supremum of this set is . We deduce this as follows. If sup{d1 , d2 , d3 , d4 } = M, then > M . Then by Corollary 3 of Chapter 1, there exists a counting number m such that 1 1 m < M and so M + m < . Now take a non-terminating decimal expansion 0 c1 c2 c3 1 1 for m . Let L be the first integer such that cL 0. Then 1 L < m . Therefore, M + 1 L < 10 10 1 < . Hence since M = sup{d , d , d , d }, d + 1 M + 1 < and this M+ m L 1 2 3 4 10 L 10 L 1 . Hence = sup{d , d , d , d } contradicts dL + L 1 2 3 4 10
his representation of the real number > 0 is unique. Suppose two non-terminating decimal expressions a0 a1 a2 a3 and b0 b1 b2 b3 are such that for some integer j 0, aj bj and ai = bi for i j 1. If a0 a1 a2 a3 represents and b0 b1 b2 b3 represents we shall show that . Suppose that aj < bj . By the hypothesis we have a j1 aj a j1 aj + 1 a1 a a1 a2 + 22 + + j1 + . j < a0 + 10 + 10 2 + + 10 j1 + 10 10 10 10 10 j a j1 a j + 1 b j1 aj + 1 a a b2 b1 But a0 + 1 + 22 + + j1 + j = b 0 + 10 + 10 2 + + 10 j1 + 10 j 10 10 10 10 b j1 bj b2 b1 b0 + + 2 + + 10 j1 + 10 j < . 10 10 a0 + Therefore, < and so . Similarly if aj > bj, we can show that . Thus this way of representing any real number > 0 is unique. Notice that for = 1 , the corresponding 2 non-terminating expansion is the expression 0.499999999999 which is a non terminating expression with recurring '9'. To complete the picture, for any < 0, we take the non-terminating decimal expansion for and represent by attaching the negative sign to the non-terminating decimal expansion for . This is consistent with the following observation. If = sup{d1 , d2 , d3 , d4 }, where the di's are as defined above in terms of the ai's and a0 a1 a2 a3 is the non-terminating decimal expansion of , then = sup{d1 , d2 , d3 , d4 } = inf{d1 , d2 , d3 , d4 }. One exception is 0 itself. It is simply represented by 0.0 a terminating decimal, that is decimal expression with recurring '0'. Cantor's Theorem. The set of real numbers strictly between 0 and 1 is not countable. Proof. The proof is a simple one, known as the Cantor's diagonal process. Suppose on the contrary that the interval {x R: 0 < x < 1} is countable. Hence there is a matching function (that is a bijective function) g: N+ {x R: 0 < x < 1} from the counting numbers N+ onto the interval {x R: 0 < x < 1}. For each counting number n let the decimal expansion of g(n) be given by 0 gn1 gn2 gn3 gn4 gn5 . (Note here that for any
55 Ng Tze Beng
Chapter 6 Genealogy
real number in the interval {x R: 0 < x < 1}, the first term of the non-terminating decimal expansion of is 0.) We then construct the following decimal expansion 0 b1 b2 b3 where bj is chosen to be different from g j j and 0 and 9. Then 0 b1 b2 b3 represents a real number > 0 and < 1 because bj 0, 9 for all j. In particular g(j) for any counting number j since bj g j j . Thus we have produced a real number in the interval {x R: 0 < x < 1} which is not in the range of g contradicting that g is bijective. This proves that the interval {x R: 0 < x < 1} cannot be countable and so the set of real numbers R is not countable. We close this chapter with the poetic remark of Eric Temple Bell: The algebraic numbers are spotted over the plane like stars against a black sky; the dense blackness is the firmament of the transcendentals.
This image sums up just how little we know about the transcendentals and hence the real numbers. Apart from the well-known transcendentals like the numbers and e, practically none is readily available; any attempt to produce one will need quite an effort both in its construction as well as its authentication.
56 Ng Tze Beng
Epilogue
The water in a vessel is sparkling; the water in the sea is dark. The small truth has words that are clear; the great truth has great silence. Rabindranath Tagore
We have constructed the real numbers and have recovered the natural numbers, integers and rational numbers in this model of the real numbers. Henceforth these numbers will take on new meanings. We have also shown that the properties of the real numbers are the essence of a complete totally ordered field. Are there other models? Yes, there are and they are constructs of quite different nature. Each model has its own merit. Cantor based his model on fundamental sequences (Cauchy sequences), where every Cauchy sequence converges, a notion that is called metric complete. Weierstrass' model based on nested intervals is another that allows us to zoom in the position of a real number on the infinite real line. The more recent model of John Conway, in terms of Conway games, has its origin in a different conceptual consideration, that of calling number a game and is a subset of a bigger field that includes infinitely small and infinitely large 'numbers'. It allows one to look forward to an extension of the real numbers to non-Archimedean non-standard numbers. Whatever models that we choose, we can give an axiomatic definition of the real numbers based on the axioms for a complete totally ordered field. Undoubtedly there is some gain in this approach, but there is a loss of familiarity with the object itself. Let us think about this creation of the real numbers in an abstract and foundational way, free from what we think numbers or indeed what natural numbers really are. Let us pause and think about the consequence of this creation. Starting from the empty set we have obtained the real numbers using entirely set theoretic construction from von Neumann's natural number system. From the real numbers we can give definitions of complex numbers, vector spaces, real and complex algebras and so on. These are the building blocks of more complicated mathematical objects which in turn are also building blocks of even more complicated mathematical objects and so on. If we accept the consistency of (Zermelo-Frankel) set theory, we can view the whole of mathematics as being built up from set theory, that is giving it a unified axiomatic basis. This view, though controversial, gives us the assured meaning of rigour, at least in a restricted portion of analysis. The question whether the whole of mathematics can be based on an axiomatised logical footing will continue to be debated well into the next century. On a note of pragmatism, the path that we have traversed in reaching the real number system is truly one of logical clarity and that is perhaps how one should view the makings of the other branches of mathematics in this way. Gdel's incompleteness theorem does restrict us from saying that everything in mathematics is provable or not provable, but it does not make the slightest blemish on the beauty of theorems such as the Fundamental Theorem of Calculus and the Euler-Poincar Index Theorem. For the question on the foundation of mathematics, a good account is given in chapter 14 of Numbers by Ebbinghaus et al and a very rich extensive historical account in Chapter 51 of Mathematical Thoughts from Ancient to Modern Times (Volume 3) by Morris Kline.
Chapter 7 Epilogue
The Archimedean property allows us to think of the real numbers as an infinite line. Thus we can think of the real numbers as a continuum. How useful is this fact? When we say the real number system is a line, what do we mean? Does the real numbers being a continuum gives us a justification to equate real numbers with the infinite line? Yes, to some extent it does. The question of what a real number is, fundamentally, is a philosophical one. To reconcile with the technicality of its construction, we think in very different ways. Natural numbers and the integers and to some extent the rational numbers are something we have learnt and used so effectively since our school days. We do not view them in a way towards a foundational understanding; we do not think of them as an object or embedded subfield in the real number system we have constructed. When we talk about the real numbers, we think of them more in terms of the properties they possessed, that is the properties of a complete totally ordered field but the embedded rational number field is not thought of in the same way. This is rightly so and we have two ways of thinking that suit our convenience. It is easy for a mathematician to switch back and forth from just plain rational numbers to embedded rational numbers in the real numbers (whether being thought of as a complete totally ordered field or a continuum). The layman would find this task difficult to say the least and the real numbers remain mysterious to fathom.
58 Ng Tze Beng
References
The books listed here meant for further readings are good source for someone who wants to know what mathematics is about. Some of them are extremely technical and some are controversial. They all have something to say about the foundation of mathematics and in particular the real numbers. They are chosen for the richness and depth of thoughts and the abundance of historical content, both inviting and challenging.
William Dunham, Journey through Genius, Penguin Books 1990. (This is a very fascinating and captivating book to begin with.) Ebbinghaus et al., Numbers, Springer Verlag 1995. ( This is a collection of the works by some prominent mathematicians on different aspects and areas pertaining to number systems and their ramification. It is a very technical book and is recommended for further reading. ) Morris Kline, Mathematical Thought From Ancient To Modern Times, Oxford University Press 1972. (A monumental work of a kind that treats the history of mathematics in a readable and thought provoking manner.) Morris Kline, Mathematics The Loss of Certainty, Oxford University Press 1980. Reinhard Laubenbacher and David Pengelly, Mathematical Expeditions, Springer Verlag 1999. (An amazing book that invites you to participate in the foot steps of some famous mathematicians.)
59 Ng Tze Beng

Real Numbers

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Real Numbers

Uploaded by

Copyright:

Available Formats

Real Numbers?

Department of Mathematics National University of Singapore 2 Science Drive 2 Singapore 117543

CHAPTER ONE Real Numbers! What are they?

Chapter 1 Real Numbers! What are they?

The Rational Numbers

Chapter 1 Real Numbers! What are they?

Chapter 1 Real Numbers! What are they?

What We Would Like the Real Numbers to Possess

Chapter 1 Real Numbers! What are they?

(Associativity) (Commutativity) (Distributivity)

Chapter 1 Real Numbers! What are they?

Chapter 1 Real Numbers! What are they?

Chapter 1 Real Numbers! What are they?

Chapter 1 Real Numbers! What are they?

Chapter 1 Real Numbers! What are they?

Chapter 1 Real Numbers! What are they?

Chapter 2 A Cut, An Instant

Chapter 2 A Cut, An Instant

Chapter 2 A Cut, An Instant

Chapter 2 A Cut, An Instant

Chapter 2 A Cut, An Instant

The Embedding of the Rational Numbers in The Real Numbers

Chapter 2 A Cut, An Instant

Chapter 3 Relearning Addition

Chapter 3 Relearning Addition

Chapter 3 Relearning Addition

Chapter 3 Relearning Addition

Chapter 3 Relearning Addition

Chapter 4 Rethinking Multiplication

Chapter 4 Rethinking Multiplication

Chapter 4 Rethinking Multiplication

Chapter 4 Rethinking Multiplication

Chapter 4 Rethinking Multiplication

Chapter 4 Rethinking Multiplication

Chapter 4 Rethinking Multiplication

Extension of Multiplication to All of the Real Numbers

Chapter 4 Rethinking Multiplication

Thus for < (0), > (0) and so

> (0) as remarked in the paragraph

6. If > (0), then > > . 7. (a) (b) = (a b).

Chapter 4 Rethinking Multiplication

Chapter 4 Rethinking Multiplication

Chapter 4 Rethinking Multiplication

Chapter 4 Rethinking Multiplication

Chapter 4 Rethinking Multiplication

Chapter 4 Rethinking Multiplication

Chapter 4 Rethinking Multiplication

Order! Order! There is only one order

Chapter 5 Order! Order! There is only one order.

Chapter 5 Order! Order! There is only one order.

Chapter 5 Order! Order! There is only one order.

Chapter 5 Order! Order! There is only one order.

1 1 1 1 1 1 1 1 (n) = (n) (1) = (n) ( (n) ( n )) = ( (n) (n)) ( n ) = (1) ( n ) = ( n ).

Chapter 5 Order! Order! There is only one order.

Chapter 5 Order! Order! There is only one order.

Chapter 5 Order! Order! There is only one order.

Chapter 5 Order! Order! There is only one order.

Chapter 5 Order! Order! There is only one order.

Proof of Theorem 9. and Theorem 15.

This is now an easy consequence of Lemma 12, Lemma 13

CHAPTER SIX Genealogy

You might also like